Documente Academic
Documente Profesional
Documente Cultură
Slagis
e-mail: slagisg@asme.org
G C Slagis Associates,
258 Hillcrest Place,
Pleasant Hill CA 94523-2184
Introduction
The design-by-analysis concept was first introduced in 1963
with the publication of the nuclear vessels code 1. In comparison
to the nonnuclear vessels code, a lower factor of safety on pressure design is incorporated. To justify the lower factor of safety,
detailed stress analysis and an evaluation of fatigue, including
explicit consideration of thermal stresses, are required. Different
categories of stress are assigned different allowable values. A criteria document 2 to explain the design-by-analysis approach was
published by ASME in 1969.
For nuclear piping, a simplified design-by-analysis approach
was first published in 1969 as USAS B31.7 3. The Foreword
section of B31.7 gives a description of the design philosophy for
nuclear piping. The piping rules were incorporated with the vessel
rules in 1971 when Section III was revised to include rules for all
nuclear components.
Some basic questions regarding interpretation of the design-byanalysis rules have come up in recent years. Some of these questions result from extensive use of finite element methods to determine stresses. For example, how are primary stresses extracted
from a finite element analysis? The purpose of this document is to
review the design-by-analysis criteria, discuss the fundamental
concepts behind the criteria, and provide insight into some of the
technical issues. The fragmented nature of code developments and
the related literature makes it difficult to fully understand all aspects of the concepts involved.
Discussions of the design-by-analysis criteria are based mainly
on the 1974 Edition of the Section III Code although the stress
limits are taken from the 2001 Code. The Code rules given in
NB-3200 apply to any pressure retaining component. The piping
rules given in NB-3600 are a simplified version of the NB-3200
rules. Some piping terms and criteria will be used to illustrate
certain aspects of design-by-analysis.
Criteria
There are two basic concepts underlying the design-by-analysis
criteria. First, stresses are categorized into three types with differ1
This is a minor revision of a paper PVP2004-2614 of the same title that was
presented at the 2004 PVP Conference.
Contributed by the Pressure Vessels and Piping Division of ASME for publication
in the JOURNAL OF PRESSURE VESSEL TECHNOLOGY. Manuscript received August 26,
2005; final manuscript received October 24, 2005. Review conducted by G. E. Otto
Widera. Paper presented at the 2004 ASME Pressure Vessels and Piping Division
Conference PVP2004, July 25, 2004July 29, 2004, San Diego, California, USA.
ent stress limits. Second, the stress limits change for different
service levels. The three categories of stress are primary, secondary, and peak. Primary stresses are load controlled; secondary
stresses are displacement controlled; and peak stresses are local
in nature. Primary and secondary stresses can be membrane or
bending.
Stress limits are established for Design, Level A, Level B,
Level C, and Level D loadings. Design conditions design pressure, design temperature, and design mechanical loads establish
the required wall thickness of the vessel. Level A conditions are
those originally referred to as normal conditions 1971 edition
and Level B as upset conditions. Level A and B loadings are
expected to occur in the operation of the component. Stress limits
for Level A and B are selected so that there is no damage to the
component that requires repair. Level C stress limits permit large
deformations in areas of structural discontinuity which may necessitate the removal of the component or support from service for
inspection or repair of damage.2 Level D stress limits permit
gross general deformations with some subsequent loss of dimensional stability and damage requiring repair, which may require
removal of the component or support from service.
The allowable limits of stress intensity from NB-3200 4 are
shown in Fig. 1 as given in the 2001 Edition. Secondary and peak
stresses are not limited for Levels C and D on the basis that
fatigue analysis is not required since only one such event is anticipated, followed by shutdown for inspection or repair 5. NB3200 Level D limits are given in Appendix F of Section III. Sm,
the allowable material stress intensity, is based on a fraction of the
material yield stress and the ultimate stress. For ferritic steels, Sm
1
2
is the lower of 3 minimum tensile strength or 3 minimum yield
1
strength. For austenitic steels, Sm is the lower of 3 minimum tensile strength or 90% of the minimum yield strength. The increase
to 90% of yield strength is to allow for the strain-hardening characteristics of austenitic steel.
When the emergency Level C and faulted Level D conditions and stress limits were first identified in 1971, the probability
of the condition occurring was discussed. For emergencyThe
conditions have a low probability of occurrence ; for faulted
Those combinations of conditions associated with extremely-lowprobability events.
Elastic stress limits for piping for emergency and faulted were
first defined in 1974. The probability of occurrence of the loads
2
Italics indicate wording taken from the Section III code document if no reference given or the referenced document.
Stress Definitions
The definitions given in NB-3213 for primary, secondary, and
peak stresses are given below.
Primary stress is any normal stress or a shear stress developed by an imposed loading which is necessary to satisfy the
laws of equilibrium of external and internal forces and moments. The basic characteristic of a primary stress is that it is
not self-limiting. Primary stresses which considerably exceed
the yield strength will result in failure or, at least, in gross
distortion. A thermal stress is not classified as a primary stress.
Secondary stress is a normal stress or a shear stress developed by the constraint of adjacent material or by self-constraint
of the structure. The basic characteristic of a secondary stress
is that it is self-limiting. Local yielding and minor distortions
can satisfy the conditions which cause the stress to occur and
failure from one application of the stress is not to be expected.
Peak stress is that increment of stress which is additive to
the primary plus secondary stresses by reason of local discontinuities or local thermal stress including the effects, if any, of
stress concentrations. The basic characteristic of a peak stress
is that it does not cause any noticeable distortion and is objectionable only as a possible source of a fatigue crack or a brittle
fracture.
The key to understanding the difference between primary and
secondary stresses is that primary stresses are required for equilibrium with an applied mechanical load. Pressure is an applied
mechanical load. The hoop stress in a cylinder to react the pressure load is a primary membrane stress. An applied moment to a
horizontal cylinder from self-weight produces a primary bending
stress. If the cylinder includes a gross structural discontinuity,
secondary stresses will also be created by a mechanical load. If
the cylinder includes a local structural discontinuity, peak stresses
will be created.
If a piping system is subjected to a fluid temperature increase,
thermal expansion stresses are created. These thermal expansion
stresses restraint of free end displacement are secondary
stresses. Thermal expansion also causes peak stress at a local
structural discontinuity a girth butt weld, for example.
A through-wall temperature gradient in a cylinder can cause a
26 / Vol. 128, FEBRUARY 2006
Failure Modes
The fundamental failure mode of concern for a pressureretaining component is burst. Another failure mode that is considered is plastic deformation. Plastic deformation yielding is a
functional concern more than a pressure boundary concern. An
owner of a vessel will not be pleased if a Level A or B loading
results in observable deformation of the vessel. From the criteria
document 2
The primary stress limits are intended to prevent plastic deformation and to provide a nominal factor of safety on the
ductile burst pressure. The primary plus secondary stress limits
are intended to prevent excessive plastic deformation leading to
incremental collapse, and to validate the application of elastic
analysis when performing the fatigue evaluation. The peak
stress limit is intended to prevent fatigue failure as a result of
cyclic loadings.
The piping equation was first adopted by the B31 Code in 1955.
Reference 7 discusses the derivation of this formula. Over 31
different formulas were considered; these formulas included elastic and plastic calculations. The final equation was selected since
it approximates satisfactorily the available room-temperature
tubular bursting data 7. The ASME B16.9 standard describes
burst test procedures for pipe fittings. Calculated burst pressure
for straight pipe is given as P = 2St / Do where S is the specified
minimum tensile strength.
Langer 8,9 discusses pressure design of vessels for burst and
provides failure pressure data on PVRC disk tests for a range of
materials. The intent of the cylindrical vessel equation, as given
above, is to provide a nominal factor of safety on burst of 3. But,
the theoretical burst pressure is dependent on the strain hardening
exponent of the material. For five different materials, the theoretical safety factor was found to vary from 2.75 to 3.34.
In NB-3200, a minimum required wall thickness is not specified
for pressure design. Instead, the primary membrane stress is calculated by elastic analysis and compared to a stress limit of Sm
for Design conditions as given in Fig. 1. Consider the limit
analysis of a simple, straight, rectangular cross-section, bar in
tension as shown in Fig. 2. Assume an elastic-perfectly-plastic
material model with a yield stress of Sy. Once the cross-sectional
stress reaches the yield stress, the maximum load carrying capability of the bar is achieved. Applying any additional load causes
the bar to deform until the failure strain is reached, and the bar
ruptures.
The normal stress in the bar is a primary membrane stress that
is required for equilibrium with the applied external load. Instead
2
1
of using Sy as the failure criterion, the code uses Sm 3 yield or 3
ultimate. Hence, the design-by-analysis criterion for primary
membrane stress provides a factor of safety of 1.5 on excessive
plastic deformation yielding. The use of ultimate tensile strength
in addition to yield strength to specify Sm accounts for the strain
hardening characteristics of the material. Ultimate failure plastic
in stability/rupture occurs when the primary membrane stress
1
reaches the ultimate strength of the material. Hence, using Sm of 3
ultimate strength provides for a factor of safety of 3 on plastic
instability or rupture failure.
By the same reasoning, the code primary membrane stress criterion provides a factor of safety of 1.5 on excessive deformation
Journal of Pressure Vessel Technology
Ductility
The design-by-analysis criteria presume ductile material behavior. Allowing secondary stresses to exceed the yield strength of
the material requires that the material have sufficient ductility to
accommodate the required plastic flow without failure. Typical
yield strains from secondary stress, and even plastic strains at
local discontinuities, are not that large in comparison to elongation to failure of 33% or more for typical carbon steels. Materials
acceptable for code use are specified, but minimum ductility is not
one of the specified parameters used for material selection. Material characteristics and ductility are discussed in Ref. 12. One
quote from this document 1964 is The amount of ductility
required to insure satisfactory performance of a pressure vessel
has never been definitively established.
There are two other aspects of the design-by-analysis criteria
that are only directly applicable if the material has sufficient ductility. For pressure design, the primary membrane stress limit is
intended to provide a nominal factor of safety of 3 on burst pressure. In a cylindrical shell with a gross structural discontinuity,
there will be significant secondary bending stresses at the discontinuity. The material must have sufficient ductility such that the
burst pressure of the cylinder is not significantly reduced.
FEBRUARY 2006, Vol. 128 / 29
Is it Primary or Secondary?
A perennial problem in running FEA is determining the primary
stress from total stress results. Judgment is definitely required.
The fallback position seems to be to consider all stresses as primary, but this is unreasonable.
The key to resolving a stress distribution into primary and secondary components is to understand that the primary stress is
required for equilibrium with an applied mechanical load. If there
is no mechanical load, or if stress is a result of compatibility
considerations at a gross structural discontinuity, the stress is secondary in nature. Consider a long cylindrical vessel with a thickness change in the middle gross structural discontinuity subjected to internal pressure Fig. 6. Simplify the structural model
by assuming an abrupt change in thickness for the stress analysis.
In this example, the thickness change is assumed to be sufficiently
removed from the vessel ends so that end conditions do not affect
the stress analysis at the discontinuity location.
Pressure causes a hoop membrane stress of pR/ t thin wall
approximation resulting in hoop strain and growth in diameter.
The growth is larger in the thinner member with the higher hoop
stress. An internal shear and moment are required to restore compatibility at the joint. Pressure is an internal mechanical load. The
hoop membrane stress is required for equilibrium with the applied
pressure primary membrane. The internal shear and moment are
self-equilibrating and are not required for equilibrium with a mechanical load. Hence, the shear and moment at the discontinuity
are secondary in nature.
Consider a second geometrya cylindrical vessel with a flat
plate closure Fig. 7. This geometry is an exception to the rule as
linearization, stress categories, example problems, and recommendations. However, there are major problems with the approach in
my opinion.
The report is oriented to proper determination of membrane
stress and bending stress without categorization as primary or secondary. For primary stresses, a separate equilibrium analysis or
plastic analysis is recommended. This may be a workable solution. But the report gives the impression that the only other solution is to consider all membrane and bending stresses at a discontinuity as primary. This is unreasonable in my opinion. A
knowledgeable engineer is able to separate primary from secondary in FEA results at a discontinuity using the principle that the
primary stress is that required for equilibrium with the applied
load. If the stress is not required for equilibrium with the applied
load, then that stress is secondary.
The report also recommends determination of P + Q at a structural element not a transition element. This is a significant
limitation. In my experience, many analyses are performed to determine thermal gradient stresses for the fatigue evaluation. The
maximum thermal gradient stress is usually in the transition element, and the P + Q stress is needed in the transition element to
determine Ke for the fatigue evaluation. The structural element
approach does not seem to be workable for fatigue damage
calculations.
Meaning of Ke
The maximum value of Ke 1 / n equal to 5 for carbon steel, 3.3
for austenitic steel has been a concern for design. For many of
the high thermal transient situations in nuclear applications, the
allowable number of cycles is very low because of the high Ke
penalty factor on the fatigue evaluation. Recalling the criteria, the
range of primary-plus-secondary stress including thermal bending
may exceed the 3Sm shakedown criterion if a penalty factor is
taken on fatigue.
Reference 14 discusses a different procedure for calculating
the fatigue penalty factor as proposed for use in the French code.
A Ke factor of 1 / n is not applied to the thermal bending stress;
1 / n is only applied to the mechanical stress. The strain concentration factor applied to thermal bending is based on the Neuber
rule. There is one statement in Ref. 14 that I do not agree with
The ASME III NB 3200 rule for Ke definition is clearly devoted to
elastic-follow-up effects as stated in Ref. 6.4 The intent of the 1 / n
factor is a critical issue.
To understand the use of Ke it is necessary to review the development of the simplified elastic-plastic method. This method was
originally developed for piping and published in 1969 in B31.7
3. The problem in Class 1 nuclear piping was that secondary
thermal gradient stresses were exceeding the 3Sm shakedown to
elastic action limit in many cases. There was no technique available to qualify the piping for fatigue without a simplified elasticplastic method. Hence, the B31.7 approach was developed. Secondary thermal gradient stresses thermal bending in NB-3200
terms could exceed 3Sm provided a penalty was taken on the
fatigue analysis.
A full discussion of the development of the simplified elasticplastic rules is given by Slagis 15. The background and technical
basis for the B31.7 approach is explained by Tagart in Ref. 16.
The B31.7 approach included two penalty factorsa notch factor
and a plastic strain redistribution factor. The notch factor accounts
for detrimental effects of plastic cycling at a stress concentration.
The plastic strain redistribution factor accounts for underestimation of strain by elastic analysis at a gross structural discontinuity
when the weaker member yields.
The 1968 edition of Section III did not have simplified elasticplastic rules to allow secondary stresses to exceed 3Sm. When the
B31.7 rules were incorporated into Section III in 1971, the sim4
for Sn 3Sm
= 1.01 n/nm 1
Sn/3Sm 1
= 1/n
for Sn 3mSm
Summary
Stresses are categorized as primary, secondary, or peak. Primary
stresses are a concern for deformation, burst, or collapse. Secondary stresses are limited to require shakedown to elastic action to
ensure the applicability of the fatigue evaluation. Peak stresses are
a concern for fatigue. Only primary stresses are evaluated for
Level C and D. The Level C and D stress limits permit large
deformations that may require repair or replacement of the component. Implicit in the Level C and D limits are lower factors of
safety against failure based on lower probability of occurrence of
the load.
Primary stresses are required for equilibrium with an internal or
external applied mechanical load. Pressure is a mechanical load
and causes primary stress. Thermal expansion in a piping system,
or a through-wall temperature distribution, is not a mechanical
load and, therefore, produces a secondary stress. A secondary
stress is displacement controlled and is self-limiting. A mechanical load can also cause secondary stresses. Stresses from internal
forces and moments, required for compatibility at a gross structural discontinuity, are secondary.
Stress limits are derived from application of limit design theory,
but limit load is not the failure criterion for primary stress. Burst
and collapse are the fundamental failure modes of concern for
primary stress.
Primary membrane stresses are limited to 1Sm for Design conditions. Sm is the lesser of 2 / 3Sy or 1 / 3Su. Cylinder burst test data
indicate that failure will occur when the hoop membrane stress
reaches the ultimate stress of the material. Hence, the primary
stress limits for Design provide for a nominal factor of safety of 3
on burst. For Level D, the nominal factor of safety is reduced to
1.43.
Primary bending stresses actually, membrane plus bending are
limited to Sm for Design conditions. For a rectangular section,
is 1.5. The limit for bending is higher than for membrane because
of the plastic hinge effect. The elastically predicted moment stress
corresponding to a fully plastic hinge is 1.5Su. Hence, an elastic
bending stress limit of 0.5Su 1.5 Su / 3 provides a factor of
safety on ultimate collapse of 3 for a design mechanical load. For
Level D, the nominal factor of safety is reduced to 1.43.
The primary-plus-secondary stress range limit of 3Sm is to ensure shakedown to elastic action of the through-wall membrane
and bending stresses. If the through-wall membrane and/or bending stresses exceed the limit, plastic cycling rather than elastic
cycling will occur. Plastic cycling at a local structural discontinuity, such as a notch, is detrimental to fatigue life. The secondary
stress limit also provides protection against ratcheting as long as
the hoop membrane pressure stress is less than 0.5Sy.
The elastically predicted primary-plus-secondary-plus peak
stress range for each unique stress cycle is used in the fatigue
evaluation. Acceptable number of cycles is determined from a
design fatigue curve. The design fatigue curve is based on best-fit
polished bar specimen data. Factors of 2 on stress and 20 on
FEBRUARY 2006, Vol. 128 / 31
cycles are used on the best-fit data curve to obtain the design
curve. The design curve is not based on crack initiation. A cumulative usage factor of 1 implies reasonable assurance that leakage
will not occur in the design life.
Simplified elastic-plastic analysis rules are provided in designby-analysis. The primary-plus-secondary stress range limit of 3Sm
may be exceeded for thermal bending provided a penalty factor,
Ke, is taken on the fatigue analysis. Thermal bending is the secondary bending from a through-wall temperature gradient or a
mean temperature difference. In piping terms, these stresses are
ET1 / 21 and C3EabaTa bTb.
The maximum value for Ke, 1 / n, is extremely conservative and
should be revised. The simplified elastic-plastic analysis method
was first developed for piping and published in B31.7. Two penalty factors were specified in the B31.7 methoda notch factor
and a plastic strain redistribution factor. This is the approach that
should be adopted for design-by-analysis. The notch factor accounts for plastic cycling at a local structural discontinuity. The
plastic strain redistribution factor accounts for underestimation
of plastic strain by elastic analysis at a gross structural
discontinuity.
NB-3200 contains code rules on Thermal Stress Ratchet. To
prevent ratcheting, limits are placed on thermal stress from
through-wall temperature distributions as a function of the value
of sustained pressure membrane stress. Ratcheting is incremental
deformation on each cycle of loading. The critical parameters are
a sustained primary stress and a cyclic secondary stress. General
loading cases can be evaluated by the Bree diagram. This is a
one-dimensional analysis based on elastic-perfectly plastic material behavior.
The design-by-analysis criteria implicitly assume materials with
sufficient ductility to accommodate the required plastic flow without failure. A measure of sufficient ductility has not been quantified to date. The burst pressure of a vessel with a gross structural
discontinuity could be significantly reduced if the material does
not have sufficient ductility. Direct application of the design-byanalysis criteria to high strength materials with low ductility is
questionable in my opinion.
One major problem in finite element analysis is separating primary stresses from secondary stresses. The analyst must exercise
competent engineering judgment. The key to making a decision is
that primary stresses are required for equilibrium with an applied
mechanical load. In general, shear and local bending moment
stresses at a gross structural discontinuity are secondary in nature.
Nomenclature
A
C3
Do
E
Eab
K3
Ke
K
m,n
Ro
Salt
Sm
Sn
Sp
Su
Sy
T
tm
T1
T2
References
1 ASME Boiler and Pressure Vessel Code Section III, 1963 Edition, Rules for
Construction of Nuclear Vessels.
2 Criteria of the ASME Boiler and Pressure Vessel Code for Design by Analysis
in Sections III and VIII, Division 2, 1969, ASME.
3 USA Standard Code for Pressure Piping, Nuclear Power Piping, USAS
B31.71969, ASME.
4 ASME Boiler and Pressure Vessel Code Section III, 2001 Edition, Rules for
Construction of Nuclear Facility Components.
5 Bohm, G. J., and Stevenson, J. D., 1982, Extreme Loads and Their Evaluation
With ASME Boiler and Pressure Vessel Code Limits, Pressure Vessel and
Piping: Design Technology1982A Decade of Progress, ASME, pp. 415
418.
6 Slagis, G. C., 1991, Basis of Current Dynamic Stress Criteria for Piping,
Weld. Res. Counc. Bull., 367, pp. 1516.
7 Burrows, W. R., Michel, R., and Rankin, A. W., 1952, A Wall Thickness
Formula for High-Pressure, High-Temperature Piping, ASME Paper No. 52A-151, p. 6.
8 Langer, B. F., 1972, PVRC Interpretive Report of Pressure Vessel Research,
Section IDesign Considerations, in Pressure Vessel and Piping: Design and
Analysis, A Decade of Progress, Volume 1, Analysis, ASME, New York, pp.
860 reprinted from Weld. Res. Counc. Bull., 95, 1964.
9 Langer, B. F., 1972, Design-Stress Basis for Pressure Vessels, in Pressure
Vessel and Piping: Design and Analysis, A Decade of Progress, Volume 1,
Analysis, ASME, New York, pp. 8494 reprinted from Exp. Mech., 1971.
10 Miller, D. R., 1959, Thermal-Stress Ratchet Mechanism in Pressure Vessels,
J. Basic Eng., 81, pp. 190196.
11 Bree, J., 1967, Elastic-Plastic Behavior of Thin Tubes Subjected to Internal
Pressure and Intermittent High-Heat Fluxes with Application to Fast-NuclearReactor Fuel Elements, J. Strain Anal., 23, pp. 226238.
12 Gross, J. H., 1972, PVRC Interpretive Report of Pressure Vessel Research,
Section 2Material Considerations, in Pressure Vessel and Piping: Design
and Analysis, A Decade of Progress, Volume 3, Materials and Fabrication,
ASME, New York, pp. 434 reprinted from Weld. Res. Counc. Bull., 95,
1964.
13 Hechmer, J. L., and Hollinger, G. L., 1998, 3D Stress Criteria Guidelines for
Application, Weld. Res. Counc. Bull., 429.
14 Grandemange, J. M., Heliot, J., Vagner, J., Morel, A., and Faidy, C., 1991,
Improvements on Fatigue Analysis Methods for the Design of Nuclear Components Subjected to the French RCC-M Code, Weld. Res. Counc. Bull.,
361.
15 Slagis, G. C., 2005, Meaning of Ke in Design-by-Analysis Fatigue Evaluation, PVP2005-71420.
16 Tagart, S. W., 1972, Plastic Fatigue Analysis of Pressure Components, in
Pressure Vessel and Piping: Design and Analysis, A Decade of Progress, Volume 1, Analysis, ASME, New York, pp. 209226 reprint of ASME Paper No.
68-PVP-3, 1968.