You are on page 1of 7

The ASME Code and 3D Stress

J. L. Hechmer Evaluation
Advisory Engineer,
Nuclear Equipment Division. TheASME Code [1] identifies the modes offailure that must be addressed to ensure
acceptable pressure vessel designs. The failure modes addressed in this paper are
precluded by limits on the primary and primary plus secondary stress. Both involve
G. L. Hollinger the transition from elasticity to plasticity. Their evaluation requires the computation
Code Specialist, of membrane and bending stresses {the linearized stresses). The original techniques
Research and Development Division.
for evaluating the limits were based on beam and shell theory. Since beam and shell
theory were the basis of the then-current tools, the transition from analysis results
Babcock and Wilcox, to failure assessment was straightforward. With the advent offinite elements (FE),
A McDermott Company, the transition from the stress distribution to the failure modes requires a different
Barberton, OH 44203
path. For three-dimensional finite element (3D FE), the path is obscure. Since the
development ofFE, theASME Code has made no additions to clarify the correlations
between FE stress distributions and the failure modes. The authors believe that the
Code should provide guidance in this area.

Introduction
Background. The ASME Code [1] places limits on the pri- 5 How much material must become plastic and what is the
mary membrane-plus-bending stress intensity (PL + Pb) and shape of the plastic zone before a plastic hinge is formed or
the primary-plus-secondary stress intensity range (P + Q) to before the material will no longer shakedown to elastic action?
preclude specific failure modes. The ASME Code failure cri- 6 Are complex structures more susceptible to failures on
teria [2] are based on the fundamental assumptions of beam- curved surfaces or in volumes of material than on flat planes?
shell theory which are that the membrane and bending stresses 7 Is the generalized von Mises yielding a more rational and
act on a plane and that planes remain plane. The Code uses acceptable failure theory for 3D FE stress distributions than
a two-dimensional shear theory of failure (plane body), where Tresca?
the three principal stresses give three stress differences, each
associated with its own plane. Thus, the current criteria require
the definition of three bending planes. Published Work
Issues. The Code limits were developed for two-dimen- As stress distributions become more complex, defining the
sional, axisymmetric geometries and loads, analyzed with shell membrane and bending stresses becomes more complex. For
theory, where bending planes can be defined and the planes 3D FE analyses, the engineer is faced with the evaluation of
remain plane. Thus, before determining the membrane and complex patterns of stress. In the Decade of Progress [3],
bending stresses, the potential failure plane is chosen. Under Kroenke et al. divide three-dimensional conditions into six
3D conditions, a single bending plane for all loads may not categories. Though these categories are arbitrary, they give an
exist or a chosen plane may not remain planar during loading. organized method of approaching the problem. In increasing
Therefore, relating finite element (FE) results to the Code limits levels of complexity, the categories are:
raises procedural issues, such as:
1 asymmetric loads on axisymmetric geometries;
1 How should membrane and bending stresses be obtained 2 perforations in otherwise axisymmetric geometries;
from 3D nonlinear stress distributions? 3 cylinder-to-cylinder intersections;
2 What can be defined as a plane when a simple definition 4 noncircular attachments to cylindrical shells;
does not exist (e.g., geometry is complex) or when planes do 5 nonradial penetrations and attachments to shells;
not remain planar? '6 general three-dimensional geometries.
3 How should bending stresses be obtained from 3D stress
distributions? As an outgrowth of Kroenke's work [3], the authors initiated
4 Is it valid to use a nonconsistent bending axis (maximum work on defining the issues. In the paper, "Three-Dimensional
versus minimum load case) in evaluating the P + Q range? Stress Criteria—A Weak Link in Vessel Design and Analysis"
[4], Hechmer and Hollinger explored the issues, with emphasis
Contributed by the Pressure Vessels and Piping Division for publication in
on the primary and secondary stress failure modes and their
the JOURNAL OF PRESSURE VESSEL TECHNOLOGY. Manuscript received by the PVP
relationship to 2D and 3D FE results. Three approaches were
Division, January 2, 1990; revised manuscript received December 14, 1990. discussed for obtaining the membrane and bending stresses for

Journal of Pressure Vessel Technology NOVEMBER 1991, Vol. 113 / 481

Copyright © 1991 by ASME


Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use
membrane and bending stresses
1— „ _ GLOBAL STRESS COMPONENTS integrated over surface or plai
I (X,Y,XY) TYPICAL POINT
Y/ / //
'///\
t \/ TYPICAL / / / \
\ \ SURFACE OR PLANE >.

" K//V//A
A V N
LOCAL STRESS COMPONENTS
(r, 1, rL)
membrane and bending sere;
integrated along a line

NOZZLE/CYLINDER INTERSECTION

PRINCIPAL STRESS "COMPONENTS" Fig. 2 Three approaches—point, line, plane


(SI, S2)
GLOBAL AXES (X,Y)
directions varly along line/plane

Fig. 1 Reference direction on a line or plane

3D conditions; stress-at-a-point, stress-along-a-line, and stress-


on-a-plane (Fig. 2). Kroenke had published [5] a procedure to
develop the membrane and bending stresses for axisymmetric
conditions. Kroenke's method develops the load distribution
from the stress distribution. He then determines the average
load and the bending moments from which the membrane and
bending stresses are obtained. The stress-along-a-line ap-
proach, discussed in [4], is based upon Kroenke's axisymmetric OUTSIDE VIEW IHSIOE VIEU
procedure, but Kroenke's procedure is only indirectly appli- Fig. 3 Three-dimensional finite element model—nozzle/cylinder inter-
cable to 3D conditions. The three approaches are defined and section
discussed in Appendix A.
The paper, "Three-Dimensional Stress Criteria—Applica-
tion of Code Rules" [6], presents a quantitative comparison opt for linearizing only two of the component stresses. The
of the application of the three approaches, using a nozzle- principal stress tensor represents three stress vectors, each
cylinder assembly. The study demonstrates that the three ap- aligned with its own set of axes, which may vary from point
proaches can give substantially different results. The most to point on the plane. Some analysts opt to use principal stresses
complex of the three approaches is stress-on-a-plane. For three- by assuming they are normal to the bending plane. Figure 1
dimensional geometries, the definition of the plane is subjective illustrates the local, global, and principal directions.
and the resultant stresses and conclusions are a function of
engineering judgment. On the other hand, the stress-on-a-plane Example Applications
approach should be the most accurate of the three approaches.
The paper, "Code Evaluation of 3D Stresses On a Plane" [7], Three example studies have been performed. The first [6]
investigates the significance of several choices in establishing compared the three approaches of stress-at-a-point, stress-
the plane. along-a-line, and stress-on-a-plane. The results give a global
view of the divergence of results, depending on choice. As a
Kroenke [5] identified a second issue, "which stresses should follow-up to this, a second [7] study zeroed-in on the problems
be linearized?" In cylindrical shells, there are four component of defining planes. This study also compares results between
stresses and three principal stresses. At discontinuities, signif- the stress-on-a-plane approach and the stress-along-a-line ap-
icant shear stresses occur and neither the shear nor radial proach. The third study [8] addressed the different options for
stresses have linear distributions. High shear stresses also cause linearizing stresses. The following paragraphs summarize de-
the principal stresses to vary in direction for adjacent points. tails on each of these studies.
Therefore, methods for obtaining the bending stress are not
obvious. As a result, there is a wide range of procedures that The General Approach (Example 1). An evaluation of a
are used for calculating bending stresses; this results in a lack nozzle-cylinder intersection (Fig. 3) was made from 3D FE
of consistent evaluation of 3D FE results. The paper, "Con- results to demonstrate the problems in using the three pro-
siderations in the Calculations Of the Primary-Plus-Secondary cedures. Two general locations were studied; one in the nozzle
Stress Intensity Range for Code Stress Classification" [8], and one in the cylinder. All three approaches are used to
addressed this issue; it presents an evaluation of seven pro- evaluate three load cases. Each load case generates represent-
cedures for determining membrane and bending stresses using ative patterns of stress which would typically be used to eval-
a simple axisymmetric analysis and Kroenke's linearization uate collapse (PL + Pb), or incremental collapse and strain
procedure. Appendix B provides additional discussion. concentration (P + Q). The evaluation shows the expected
A common approach to obtaining bending stresses is to wide variation of results on even a simple geometry subjected
determine an effective moment, based on the load distribution. to simple but representative loadings.
The technique uses component stresses to obtain the loads, For the stress-along-a-line procedure, the component stresses
because they represent vectors, all of which are consistent to were linearized along a line to obtain six membrane-plus-bend-
a specific set of axes (global or local). Two of the component ing component stresses. For the stress-on-a-plane procedure,
stresses will be normal to a bending plane; thus, many analysts different planes were evaluated to show the impact of this

482 / Vol. 113, NOVEMBER 1991 Transactions of the ASME

Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Table 1 Summary of Results Table 2 Summary of findings (use of SCL versus plane)
Stress Intensities (ksi) LOADING AND STRESS CATEGORY
Case Description Point Line Plane Pressure Pressure Thermal Thermal
Plane membrane bending membrane bending
1 Nozzle pressure- 37.7 > 32.9 > 32.3-28.6 Ring Signifi- Invalid ap- Invalid appli- Invalid appli-
only cantly un- plication
2 Shell pressure- 43.6 > 42.3 > 40.9-35.0 cation cation
only conserva-
3 Nozzle pressure 54.7 > 32.3 > 29.9 tive
+ thermal'"' Nozzle Good for Good Good Good for
4 Nozzle pressure 60.7 > 32.9 > 29.9 small- small
+ thermal range planes— planes—un-
5 Shell pressure + '47.8 > 42.9 > 41.0-34.1 becomes conservative
thermal conserva- for large
6 Shell pressure + 60.7 > 42.9 > 41.0 tive for planes
thermal range large
'"'Thermal stress calculated by imposing nodal-point temperatures planes
computed from a typical corresponding time-dependent finite element Fillet Good, but Good for Good Good, but
thermal analysis of the model. increas- small potentially
ingly con- planes— unconserva-
servative overly con- tive
servative for
large planes
Shell Good Good for Conservatism Unconserva-
small could be tive
planes— large
overly con-
servative for
'H" MOZZLE PLANE large planes

plane; its location, extent, shape, and orientation. It is im-


portant to encompass the area that represents a plane-of-bend-
*R" MUG PLI.KE ing whereon {P + Q) stresses are a measure of the structure's
\'• '•"U-"l
*S" SHELL PLANE response to shake-down loads, incremental collapse, and con-
Fig. 4 Three-dimensional evaluation planes—ring, nozzle, fillet, shell
centrated inelastic strain. An analysis was made, using the
nozzle-cylinder geometry, to study the effects of varying the
plane definition. It also compared results to the stress-along-
a-line method. A method developed by Hsu [9] is used to
arbitrary decision; each plane was established by varying the calculate the bending stresses on the planes.
circumferential dimension. The membrane stress is readily cal- Four sets of evaluation planes are used; they are designated
culated on a plane, but bending stress on a plane can be ob- RING, NOZZLE, FILLET, and SHELL as shown in Fig. 4.
tained in many ways. For this initial evaluation, two sets of The "points 1, 2, 3, 4" define corners of the planes. Each of
stress-along-a-line results were used to determine the bending the basic planes is studied with variations in dimension. The
stresses on the plane. lines from points 1 to 2 are used as the "reference" stress
Table 1 provides a summary of results, from which the classification line (class-line) for comparison of results with
following observations are made: the stress classification planes (class-planes).
The evaluation planes are chosen to illustrate the possible
1 The difference of results between the three methods is variation of (P + Q) stresses computed for typical locations
significant, particularly in Case 4. in a typical nozzle-cylinder assembly. The choices are limited
2 The degree of apparent conservatism decreases in order and designed to show the trends in computing stresses on planes
from stress-at-a-point, line, then plane. of various dimensions and their corresponding "reference"
3 The most significant differences occur between stress-at- class-lines.
a-point and stress-on-a-line. The differences between the stress- This study is qualitatively summarized in Table 2. The fol-
along-a-line method and stress-on-a-plane method are less sig- lowing conclusions were reached:
nificant.
4 There can be a wide variation of results for evaluation 1 Class-plane stresses converge to class-line stresses as the
of stress-on-a-plane. This indicates the significant effect of class-planes are made smaller, indicating that the method is a
choosing various planes and the potential nonconservatism "valid" extension of the stress-along-a-line method.
associated with some choices. 2 The relative conservatism between the two methods is
mixed; i.e., neither method always gives the most conservative
Though not shown in Table 1, the stress-at-a-point method results.
sometimes predicted a lower (P + Q) stress than the other 3 For evaluation of the ring, the plane encompassing the
two methods. This serves to illustrate that the "banner-of- , entire ring cross section is required to obtain meaningful re-
conservatism" often carried for the "stress-at-a-point" method sults.
is not always justifiable. This is caused by the total stress on 4 For nozzle planes, the class-line does a good job of pre-
one or both of the surfaces being less than the equivalently dicting the membrane and bending stresses for both pressure
linearized stress predicted by the stress-on-a-line or plane meth- and thermal load. For pressure membrane load, the 45-deg
ods. It typically occurs at locations of nonlinear gradients of plane gives 10 percent lower stresses than the smaller planes.
stress. For thermal bending load, the stress increases for planes greater
than 30 deg, making the class-line approach unconservative.
Definition of Plane (Example 2). The computation of (P 5 The results are mixed for the fillet planes. For pressure
+ Q) stresses on a flat plane begins with the choice of the loads, the class-line is conservative, but reasonably accurate

Journal of Pressure Vessel Technology NOVEMBER 1991, Vol. 113 / 483

Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


for planes up to 30 deg; then it becomes overly conservative. different methods give substantial differences in results; i.e.,
For thermal loads, the class-line is good for membrane loads, the choice of method is significant. However, on the basis of
but unconservative for bending loads. a single geometry, it is difficult to make judgments on which
6 In the shell locations, the class-line results are mixed; procedure is best. The evaluations of the various considerations
they are overly conservative for pressure bending stresses, and give a number of indications.
they are unconservative for thermal bending stresses.
7 This study shows that the choice of size of the plane is 1 It is logical to linearize only the stresses that are normal
important. to a plane.
2 When not linearizing a component or principal stress
Linearization Procedures (Example 3). The geometry is a type, it is reasonable to use the total value at the surface.
cylinder-to-plate assembly (Fig. 5). Two locations within the 3 It seems appropriate to linearize the principal stresses
discontinuity were evaluated; these are designated class-line 1 for cases where the stress distributions do not support the use
and class-line 2. Neither of these is an ideal plane. Class-line of total component radial or shear stresses.
1 is probably the better representation through the majority 4 The linearization choices may vary depending on the
of the thickness, but it is not normal to the inside surface. location and orientation of the assumed bending plane.
Class-line 2 is normal to the inside surface, but not to the 5 For this geometry and loading, the best approach appears
outside surface. Therefore, the analysis addresses a condition to be the linearization of the two normal principal stresses and
for which an ideal bending plane cannot be defined, but where the use of the surface values for the through-thickness principal
membrane-plus-bending stresses are required for evaluation of stress.
strain concentration.
Actually both class-lines did a good job of producing stress
distribution that were close to their expected patterns. Table Other Failure Theories
3 summarizes the results. The ASME Code uses Tresca or the shear theory of failure;
For all four combinations, the hoop (6) and the meridional it is straightforward and simple to apply for axisymmetric
(Z) stress distributions are linear. Also, the meridional prin- conditions. There are other failure definitions that can be used
cipal stress distribution was linear, except for "class-line 2 to relate 3D FE stress distributions to the Code failure modes.
thermal" where it is judged as slightly erratic. The shear stress They may offer improvements in accuracy; however, they have
distribution is generally linear, but it should be parabolic. The their own weaknesses. The fundamental question is defining
radial stress distribution is parabolic, but it should be linear the onset of a plastic hinge or shakedown. The first of these
for the pressure cases. Class-line 2 does get the correct stress methods is a 3D shell theory approach.
level at the surface. The surface principal stress, S2, is correct.
Table 3 leads to the conclusion that linearizing the 0-stress Shell Theory. Plate and shell structures exhibit their own
(which is a principal stress for this geometry) and the SI stress fundamental membrane and bending stress interactions. WRC
and using the surface values of the S2 stress give good results. Bulletin 250 [10] provides a more accurate method than beam
Table 4 presents a summary of the numerical results; the six theory for computing the membrane and bending stress inter-
methods are defined in Appendix B. It is obvious that the action in plate/shell structures. It uses the fundamental plate/
shell internal forces and moments in the two localized principal
directions of a plate/shell differential equilibrium element. The
AXISYMMETRIC FINITE ELEMENT MODEL Tresca failure theory is used and the method computes an
interaction failure locus consistent with the Code's primary
stress intensity limits of Sm and \.5Sm.

s
Table 3 Summary of linearization results
Class- Load case e R Z RZ SI S?
line
1 Pressure L P L L L F*
ft- 2
Thermal
Pressure
L
L
P
p*
L
L
P
L
L
L
P*
P*
Thermal L P L L E P*
L = linear (with or without concentration)
P = parabolic
P* ~ parabolic with correct surface values
Fig. 5 Axisymmetric finite element model—cylinder-to-plate lineari- E = erratic
zation study E* = erratic with correct surface values

Table 4 Stress difference SVL


SCL 1 SCL 2
Pressure Thermal Pressure Thermal
Method Ins Out Ins Out Ins Out Ins Out
1 20.0 7.0 30.1 28.7' 18.6 7.9 22.0 25.5
2 22.8 7.0 35.5 28.9 19.0 8.3 22.2 26.3 (w
3 18.2<o) 8.2 29.1 28.9 18.0(o) 7.1"" 20.4(<" 21.9<0)
6)
4 24.1< 6.9<0) (i
37.1 " 28.4(<" 21.3"" 7.5 26.8<6) 24.6
5 19.2 9.1 (6) 31.7 31.6(*> 18.7 7.7 24.2 25.4
6 20.3 7.1 27,1(6) 29.7<4) 18.9 8.4(i» 23.2 26.1
Percent
Max - Min 24 24 27 10 15 15 24 17
Max
'"'minimum values
""'maximum values

484 / Vol. 113, NOVEMBER 1991 Transactions of the ASME

Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Bench-marking tests using WRC 250 with existing designs to incorporate guidelines for relating finite element results to
show a mixture of conservatism and nonconservatism when the failure criteria. The total process may require research into
compared with current Code beam interaction methods. This the adaptability of beam theory versus the development of
mixture of results is not an indication that the method is un- alternate criteria. Only through consistency in application of
usable. Rather, it is a more accurate method from which dif- the tools is the "design by analysis" philosophy credible. Work
ferent results are to be expected. is underway in PVRC [13] to develop consensus of direction
One shortcoming to using this plate/shell interaction concept for development of guidelines for application to the current
in its current state of development, is that it has not yet been Code and for development of direction for future work, per-
formulated for secondary-type loadings such as thermal loads. haps to modify the Code.
Therefore, an essential step in pursuing its use in the Code is
its formulation for shakedown. References
1 ASME Boiler and Pressure Vessel Code, Section III, "Rules for Construc-
Distortion Energy Theory of Failure. Another approach tion of Nuclear Power Plant Components," Div. 1, Subsection NB, Class 1
might be to allow the von Mises failure criteria as an option Components, and Section VIII, "Rules for Construction of Pressure Vessels,"
to Tresca. Although Tresca is more conservative than von Div. 2-Alternative Rules, The American Society of Mechanical Engineers, New
Mises, this should not deter us. The Criteria document [2] also York, NY, 1986 Edition.
2 Criteria of the ASME Boiler and Pressure Vessel Code for Design by
notes the greater simplicity of using shear theory over distortion Analysis in Sections III and VIII, Div. 2, ASME, 1969.
energy. The document states on p. 3 in the last part of the 3 Kroenke, W. C , et al., "Component Evaluation Using the Finite Element
first paragraph under the "Strength Theories" heading: "Most Method," Pressure Vessel and Piping Technology—1985—A Decade of Prog-
experiments show that the distortion energy theory is even more ress, The American Society of Mechanical Engineers.
4 Hechmer, J. L., and Hollinger, G. L., "Three-Dimensional Stress Cri-
accurate than the shear theory, but the shear theory was chosen teria—A Weak Link In Vessel Design and Analysis," PVP Vol. No. 109, A
because it is a little more conservative, it is easier to apply, Symposium on ASME Codes and Recent Advances in PVP and Valve Tech-
and it offers some advantages in some applications of the nology Including a Survey of Operational Research Methods in Engineering,
fatigue analysis." Today, post-processing is so computer-ori- eds., J. T. Fong, et al., The American Society of Mechanical Engineers, New
York, NY, July 1986.
ented that there is no difference to the designer and very little 5 Kroenke, W. C , "Classification of Finite Element Stress According to
to the programmer. A strong case can be made that finite ASME Section III, Stress Categories," The Winter Annual Meeting of The
element output is much more compatible with distortion energy American Society of Mechanical Engineers, Nov. 1973.
approaches than shear stress theory; however, there are also 6 Hechmer, J. L., and Hollinger, G. L., "Three-Dimensional Stress Cri-
problems in this approach. teria—Application of Code Rules," PVP-Vol. No. 120, Design and Analysis of
Piping, Pressure Vessels, and Components, eds., W. E. Short II, et al., The
There are two cases where distortion energy theory has been American Society of Mechanical Engineers, New York, NY, July 1987.
chosen over the shear theory. The first is the elevated tem- 7 Hechmer, J. L., and Hollinger, G. L., "Code Evaluation of 3D Stresses
perature creep fatigue rules in Code Case N-47 [11]; these are on a Plane," PVP-Vol. No. 161, Codes and Standards and Applications for
Design and Analysis of Pressure Vessel and Piping Components—1989, eds.,
based on the calculation of strain-ranges by equations equiv- J. P. Breen, et al., July 1989, The American Society of Mechanical Engineers,
alent to the distortion energy equations. Therefore, we have New York, NY.
Code acceptance of the distortion energy approach in the eval- 8 Hechmer, J. L., and Hollinger, G. L., "Considerations in the Calculations
uation of fatigue. The second is the definition of yielding in of the Primary-Plus-Secondary Stress Intensity Range for Code Stress Classi-
fication," PVP-Vol. No. 136, Codes and Standards and Applications for Design
most finite element plasticity programs. Stresses or strains are and Analysis of Pressure Vessel and Piping Components, eds., G. L. Hollinger,
calculated by distortion energy equations for comparison to et al., The American Society of Mechanical Engineers, New York, NY, June
the material yield strength, followed by changing the modulus 1988.
of elasticity of the yielded material. Therefore, a Code plastic 9 Hsu, K. H., and McKinley, D. A., "SOAP—A Computer Program for
Classification of 3D Finite Element Stresses on a Plane," PVP-Vol. No. 185,
analysis would probably use distortion energy theory to define Finite Element Analysis, Computer Applications, and Data Management, ed.,
the onset of plasticity. K. H. Hsu, The American Society of Mechanical Engineers, New York, NY,
The definition of a bending plane has some problems when June 1990.
using distortion energy. It is appropriate to conclude that in 10 O'Donnell, W. J., et al., WRC Bulletin 250, "Generalized Yield Surface
for Plates and Shells," July 1979.
the development of a three-dimensional stress criteria, the dis- 11 The ASME Boiler and Pressure Vessel Code, Code Cases, Nuclear Com-
tortion energy theory should be given in-depth consideration. ponents N-47-21 Class 1 Components in Elevated Temperature Service, 1983.
12 Gerdeen, J. C , WRC Bulletin 254, "A Critical Evaluation of Plastic
Limit Load. Gerdeen in WRC Bulletin 254 [12] discusses Behavior Data and a Unified Definition of Plastic Loads for Pressure Com-
the state-of-the-art of limit analysis as applied to pressure ponents," Nov. 1979.
13 PVRC Grants 89-16and 90-13, "Three-Dimensional Stress Criteria," work
vessels. He demonstrates that quite a bit of knowledge is avail- in progress.
able for determining the limit load for a few geometries, sub-
jected to specific loads. These are quite useful in evaluating
failure analysis procedures, especially for primary load limits. A P P E N D I X A
However, there are neither sufficient geometries covered nor
combinations of load cases evaluated to consider the limit load
The Three Methods of Stress Classification
approach suitable for design. For example, a nozzle-shell junc-
ture must be designed for six pipe loads, pressure and thermal Three methods are used for evaluating Code limits on PL
loads, all superimposed. Such complexity is beyond the current + Pb and P + Q stresses: stress-at-a-point, stress-along-a-
application of limit analysis to design. line, and stress-on-a-plane.
Stress-at-a-Point. Stress-at-a-point is the simplest of the
three methods to apply, but it is also the least accurate. The
Conclusions method consists of comparing the stresses at a single location
The evaluation of Code (P + Q) failure criteria from 3D (a point) to the Code limit. The stress-at-a-point method can
finite element results requires extensive understanding of the be used with simple geometric shapes, subjected to simple
technology. The Code and the open literature provide minimal loads; its application assumes that the state-of-stress at a point
guidance. Starting with the stress results of a 3D finite element defines the local membrane and membrane-plus-bending con-
analysis, many significantly different Code evaluations of PL ditions. Obviously, the stress-at-a-point can only define the
+ Pb and P + Q stresses can occur, depending on the analyst's membrane-plus-bending if there is no nonlinear stress distri-
persuasion. Publication of various examples is needed before bution. As geometric features and loading conditions become
an industry consensus can be reached. Then the Code needs more complex, the use of stress-at-a-point becomes inaccurate

Journal of Pressure Vessel Technology NOVEMBER 1991, Vol. 113/485

Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


and more remote from truly evaluating the Code's failure respectively. The linearization procedure (for membrane plus
modes. bending stresses) affects the surface shear stress, which affects
The stress-at-a-point is usually accepted as conservative, be- the varying principal stresses.
cause the total stresses are usually much greater than equivalent The Code describes bending stresses as being applicable only
linearized stresses. When used under complex conditions, the. to normal stresses; i.e., by omission, not shear stresses. This
claim of conservatism is often used to justify the inaccurate, is addressed in paragraphs NB-3213.6, "Membrane Stress";
albeit simple method. However, cases have been found where NB-3213.7, "Bending Stress"; and NB-3213.4, "Normal
the method is not conservative. Moreover, assuming that con- Stress." In paragraph NB-3215, "Derivation of Stress Inten-
servatism is assured, the potential degree of ultra-conservatism sities," the Code describes the calculation of stress intensities
could force unfounded judgment on the acceptability of a with respect to the v'arious categories. A note is provided which
component. Therefore, stress-at-a-point should only be used states: "Membrane stress intensity is derived from the stress
where judgment indicates that the conditions are relevant to components averaged across the thickness of the section. The
the failure modes. averaging shall be performed at the component level." Since
the note refers to component stresses, the Code appears to
Stress-AIong-a-Line. In an axisymmetric shell-of-revolu- exclude the linearization of principal stresses.
tion, a line through a thickness represents a surface generated Under 3D conditions, algorithms can be developed for cal-
by revolving the line about the component's axis of revolution. culating a bending moment for each of the six component
In such a case, the circumferential distribution of stress is stresses or three principal stresses. The key issue is whether a
constant and the same as that defined by the line. Therefore, plane normal to the component or principal stress can be de-
the line used in axisymmetric analyses is actually a plane and fined and whether this is necessary for the calculation of a
stress-on-a-plane is the method used implicitly with axisym- bending moment. To visualize the condition, an incremental
metric conditions. element is used with component stresses (Fig. 6). The following
Extending the stress-along-a-line method to three-dimen- relates this geometry to each of the six stress components.
sional cases seems logical, but the line is not directly or nat-
urally relatable to any plane. As conditions become less Hoop and Axial Component Stresses. Both the hoop and
axisymmetric (structures or loading), the stresses along a line axial component stresses act perpendicular to two opposite
become similarly less related to shell-of-revolution stresses. planes of the incremental element. Their stress distribution can
Assuming a local coordinate system can be chosen, the method vary biaxially on their planes. Therefore, these planes are planes
mathematically develops membrane and bending stresses, of bending in the classical bending sense. There are no valid
through the thickness of a shell, which can be used for Code arguments against determining bending for the axial stresses.
evaluations rather than using a localized point. However, the However, in shell-type geometries, the hoop stress distribution
need for a reference plane of bending raises questions on the is actually an inverse function of the radius, even for membrane
relation of the method to predicting failure. The central issue load conditions. Thus, the normal distribution is nonlinear
is whether stress along a line, rather than on a plane, can be and it can be argued that the distribution is not indicative of
related to limit load or shakedown. a bending moment. Except for very thick shells, the nonmem-
brane level is trivial and a linear distribution is an appropriate
Stress-on-a-Plane. The computation of membrane and engineering approximation. In many 3D geometries, there is
bending stresses on a planar surface is the most complex and no true circumferential condition, but there will be two stress
potentially the most accurate method. The mathematics are components that act normal to the incremental element, with
similar to that used in the stress-along-a-line method in axi- variable stress distributions.
symmetric analyses. Actually, the mathematics are an exten-
sion of the stress-along-a-line method. The method uses Radial Component Stress. When the incremental element
integration of stress to get the load distribution, averages the is oriented so that a component stress is truly radial (through
load to get the membrane stress, and calculates the maximum thickness) in a constant thickness shell, the radial component
bending moment about orthogonal axes on the plane. Thus, stress is normal to two planes of the incremental element.
the algorithm to compute membrane and bending for each However, the radial stress is always constant on its plane. The
component stress can be written. stress distribution varies through the thickness which is a basis
Even though stress-on-a-plane is the most general method for calculating a moment for this component. Thus, bending
of evaluation, significant difficulties are encountered in choos- can be calculated, but has no physical meaning. The radial
ing location, orientation, and the extent of the plane, even stress in a constant thickness shell is usually small (relative to
more-so than choosing a line in the stress-along-a-line method. hoop and axial) and whether or not a bending stress is cal-
For axisymmetric conditions, the engineer must pre-suppose culated will have a minor impact on the results.
a set of three planes of failure which relate to plastic collapse
and progressive distortion, but the planes are well established
by the axisymmetric geometry. The problem for 3D geometries
is that the choice of bending planes is not necessarily obvious;
i.e., the orthogonal coordinate system that will establish the
bending stresses can change with load cases. Therefore, con-
siderable judgment or testing is involved.

APPENDIX B

Approaches for Determining the Bending Stress


The ASME Code [1] addresses the calculation of P + Q
stresses in paragraph NB-3222.2 with footnote reference to
Paragraph NB-3216. The footnote reference mentions alge-
braic signs of principal stresses and implies reference to the
two subparagraphs to NB-3216, NB-3216.1, and NB-3216.2
which address constant and varying principal stress directions, Fig. 6 Incremental element

486 / Vol. 113, NOVEMBER 1991 Transactions of the ASME

Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use


Within discontinuities, planes for stress classification are are based on the nonuniform direction of principal stresses;
never both radial to the surface and normal to the meridional i.e., the direction of the principal stresses can be different at
direction. As a result, the radial stress distribution (parallel to each point on the classification plane. As yet unknown, is
the plane of stress classification) is actually a combination of whether using the principal stresses is the same as using all six
radial and axial stress. Another way to say this is that the axial component stresses before calculating the principal stresses.
stress direction does not match the meridional direction, where In evaluating these choices, one must keep in mind that if the
the meridional stress is the stress of concern. As a result, the plane is a true bending plane and if it remains plane, both the
correct meridional bending stress is obtained when the radial component stress approach and the principal stress approach
stress is linearized. Therefore, true radial stresses have no phys- should give comparable results.
ical relation to bending, but what is called the radial stress To have a consistent principal stress distribution, the normal
may be skewed and not truly radial and thus contain a bending component stress distribution must be basically linear and the
component. shear stress distribution must be parabolic or very low. The
main problem is to define a true bending plane. At discontin-
Shear Component Stress. The shear stress does not act uities, it is often impossible to obtain a plane for bending
perpendicular to planes of the incremental element; therefore, where the stress distributions are appropriate over the entire
the concept of "shear bending" has no physical meaning. Also, plane.
the normal distribution for a shear stress is parabolic, not Based on the foregoing discussion, seven options have been
linear; again, shear bending has no meaning. On the other defined for obtaining the membrane plus bending stresses for
hand, cases occur where there is a nonparabolic shear stress use in PL + Pb and P + Q evaluations:
distribution. This indicates that the axial stress does not rep- 1 use all six component stresses;
resent the true meridional stress; i.e., the plane chosen for 2 use the three normal component stresses and use the total
stress classification does not represent the critical bending plane. shear stresses at the surface;
Therefore, to develop a principal stress that coincides with the 3 use the three normal component stresses and use the mem-
critical vector, the shear must be linearized. The only other brane shear stresses;
option would be to choose a better plane of stress classification. 4 use the two normal stresses that act on a plane (nominally
the hoop and meridional), use the total for the third normal
Principal Stresses. In lieu of using component stresses to (radial) stress, and use total shear stress at the surface;
determine the bending moment, the principal stresses can be 5 use the two normal stresses that act on a plane and use the
used, because they are component stresses in the coordinates membrane radial and shear stresses;
of zero shear stress. The main argument for using principal 6 use the three principal stresses;
stresses is that the question of "shear bending" is by-passed. 7 use the two principal stresses that act on a plane and use
On the other hand, arguments against using principal stresses the total for the third (radial) principal stress.

Journal of Pressure Vessel Technology NOVEMBER 1991, Vol. 113 / 487

Downloaded From: http://pressurevesseltech.asmedigitalcollection.asme.org/ on 01/28/2016 Terms of Use: http://www.asme.org/about-asme/terms-of-use

You might also like