Annex 6

Pressure components fatigue design

in the framework of Directive 97/23/EC on
pressure equipment

Work Package 6:

“Assembly of information on European

Standards development (CEN) relevant to
pressure equipment fatigue design”

FINAL REPORT
(draft, July 2001)

FDS DOC (2001) 06 (rev.3) Annex 6 / 297

 ASSEMBLY OF INFORMATION ON EUROPEAN
STANDARDS DEVELOPMENT (CEN) RELEVANT TO
PRESSURE EQUIPMENT FATIGUE DESIGN

Summary:

This report is a detailed review of the content and background of the fatigue design rules contained in
the European standards for pressure equipment and more specially in clauses 18 and 17 of prEN
13445-3, Part 3 of the future European standard for unfired pressure vessels. A comparison with other
national codes is carried during the review.

Although the quality of the work performed by the European experts is widely recognised, some
modifications are still necessary to foster the use of the fatigue rules contained in the European
standards by the European industry. Previous work performed under Tasks 1 to 5 of the Contract Nr
32.0186 of the European Commission and the previous review suggest the following recommendations:

- Improvement of the fatigue damage assessment by additional research,

- Clarification and simplification of the present technical requirements,

- Publication of criteria concerning the rules and worked examples,

- Inclusion of a clause in the standards on experimental fatigue analysis,

- Extension of the analysis to cover specific items not already covered, such as random loading or
fatigue behaviour under specific environments.

FDS DOC (2001) 06 (rev.3) Annex 6 / 298

Introduction

In its Annex I, the pressure equipment Directive 97/23/EC [1] requires that “the design must take
appropriate account of all foreseeable degradation mechanisms (e.g. corrosion, creep, fatigue)
commensurate with the intended use of the equipment. Attention must be drawn, in the instructions
referred to in section 3.4, to particular features of the design, which are relevant to the life of the
equipment, for example:

- for creep: design hours of operation at specified temperatures,

- for fatigue: design number of cycles at specified stress levels,

- for corrosion: design corrosion allowance.”

The European standards in preparation aim to answer to these requirements. In particular, long
developments are devoted to fatigue design in clauses 17 and 18 of Part 3 of prEN 13445 [2], the draft
European standard for Unfired pressure vessels and to a certain extent in the European standards for
Water tube boilers and Shell boilers, respectively in Part 3 of prEN 12952 [3] and 12953 [4], and in Part
3 of the standard for Metallic industrial piping prEN 13480 [5]. In a first part, this report is a review of the
content, origin and background of these rules. Simultaneously a comparison with other national codes
or standards is performed.

As the aim of this study is to foster the acceptability of European standards by the European industry,
some suggestions of improvement will be presented in the second part of this report, based on the
findings of Tasks 1 to 5 of the Contract and also on the review carried out in the first part.

1) Detailed review of the content, origin and background of fatigue design rules
of European standards
The newest developments on fatigue analysis are contained in Part 3 of the draft European standard for
pressure vessels [2]. These are the result of a long evolution that started in 1976 with the publication of
the first edition of the British Standard BS 5500 [6], where for the first time fatigue analysis made
reference to the behaviour of welded zones. This approach differs from the classical approach of the
ASME Code, which will be presented first. Then the fatigue analysis contained in prEN 13445-3 will be
analysed in detail. Finally, the fatigue analysis in other European pressure equipment standards will be
described briefly.

1.1) Background

A detailed presentation of the background to the present fatigue rules can be found in the paper
presented by Alain Handtschoewercker at the AFIAP Conference in 1995 [7].

1.1.1) The classical approach

Fatigue analysis of the ASME Code is used almost everywhere and can be considered as the classical
approach. It was introduced in Section III (nuclear components) [8] in 1963 and in Section VIII Division 2
[9] in 1968. Then it was adopted in many codes, such as the French nuclear Code RCC-M [10], the
English, Swedish, Dutch, Australian, Japanese codes, and with important differences in AD-Merkblatt
S2 up to the 1988 edition [11]. This approach is still used in the draft European Standard on Water tube
Boilers [3].

ASME approach is characterised by:

- The use of fatigue curves based on tests of smooth specimens exempt of all perturbations due
to welding.

- The use of total stresses or notch stresses according to the usual terminology, i.e. stresses
which include the linear and non-linear part of stress distribution across the thickness.

FDS DOC (2001) 06 (rev.3) Annex 6 / 299

1.1.1.1) Fatigue curves

The curves given respectively in the ASME Code and AD-Merkblatt S2 (1988 edition) are noticeably
different and the safety coefficients are also different. They have the following values:

- 2 on the stresses and 20 on the number of cycles, these values being already integrated in the
design curves of the ASME Code;

- 1,5 on the stresses and 10 on the number of cycles, these values to be applied to the fatigue
curves given, which are curves to failure, the failure being the onset of an observable crack.

In the German rule, a mean stress correction must be applied to the stress range (Gerber rule) when
the stress is not purely alternate, while the ASME fatigue curves incorporate the maximum possible
reduction due to the presence of a tensile mean stress (Goodman rule), mainly to cover the effect of
residual stresses. The resulting correction modifies the curves beyond about two millions cycles. More
details on that correction may be found in the Criteria of the ASME Code [12]. The ASME and AD-
Merkblatt (edition 1988) fatigue curves are represented in Figure 1.

1.1.1.2) Notch stress determination

As far as the determination of stresses is concerned, the problem is the same for the two Codes, since
the notch stresses must be determined. The computation must take into account the effect of any
discontinuity source of stress, whatever is the scale. In practice, two possibilities exist for the designer:

a) Shell and plate model

The designer determines the notch stress from a model based on shell and plate theory, using for
instance some finite element software. In this case the stresses are linearly distributed through the
thickness, stress concentrations due to microscopic discontinuities (notches, small radii, weld toe
transitions, etc.) being ignored. To take them into account, it is necessary to apply stress concentration
coefficients K t (theoretical stress concentration factor from the literature) or Keff (effective stress
concentration coefficient) to stresses computed adjacent to discontinuities.

When the local geometry is well defined (e.g. machined radii with known dimensions), it is possible to
find such coefficients in specialised books, such as references [13] and [14].

On the contrary, when the geometry has a random character (notably in the case of weld toes in the as-
welded condition, for instance), the selection of an appropriate coefficient is far more subjective, and
may lead to diverging interpretations.

For such cases, the ASME Code gives just maximum values of the stress concentrations to apply:

- 5 for any notch,

- 4 in case of partial penetration fillet welds.

These coefficients are called “fatigue strength reduction coefficients” in the ASME Code, and are in fact
equivalent to the effective stress concentration coefficients Keff , except that they are defined and
applicable for any fatigue endurance, whereas Keff normally refers to the endurance limit.

On the contrary, the AD-Merkblatt S2 (1988 edition) rules define values of the coefficients of stress
concentration to apply to welded joints, as a function of the type of assembly.

For this purpose, the various assemblies are classified in three groups (designated by K1, K2 and K3),
according to the notch effect, which is small, medium or high. To each of them is associated an effective
value of the stress concentration (designated by fK1, fK2, f K3) , which depends on the tensile strength of
the material under consideration and the number of stress cycles applied.

FDS DOC (2001) 06 (rev.3) Annex 6 / 300

The so-defined fK values are strictly equivalent to the fatigue strength reduction coefficients used in the
ASME Code, but take lower values than the maximum values given in the ASME Code. fK values only
reach the ASME Code values close to the endurance limit (N > 2. 106 cycles), and only for high strength
steels.

b) Refined model

In practice, the designer may use a refined analysis model with solid elements reproducing the exact
geometry of the zone under study. In this case, no additional stress concentration coefficient is
necessary, the notch stress being obtained directly. But the same uncertainty remains as regards the
local geometry to use to model random notch effects like welds.

Except in particular cases, experience has shown that such models offer no advantages over simpler
shell and plate models. Consequently, they are only used when the latter are not practical (e.g.
structures having parts of heavy thickness, rapid thermal transients).

1.1.1.3) Stress orientation

A final aspect to underline in the classical approach, which is common to the ASME Code and AD-
Merkblatt S2, is the fact that the stress direction with respect to the orientation of potential fatigue
cracking is not considered.

In these rules, fatigue design is based on equivalent stresses, or more precisely the equivalent stress
ranges. These are scalar quantities, computed according to Tresca criterion.

The use of the equivalent stress range is valid for unwelded zones, without defects or cracks, for which
the material initially presents the same resistance in all directions. This isotropy of the behaviour does
not exist in welded zones, where the fatigue strength in the longitudinal direction is frequently greater
than that in the transverse direction.

To give the best results, the fatigue analysis must take account of the stress orientation and provide
fatigue curves specific to each load direction. The classical approach offers neither of these two
possibilities.

1.1.2) Contribution of BS 5500/Appendix C (1976 to 1985 editions)

The first pressure vessel code, which referred to the fatigue behaviour of welded material, was British
Standard BS 5500 [6].

From the first edition in 1976, this Standard has provided its own fatigue curve, different from those of
the ASME Code, while the rest of the rules remained essentially in conformity with the American Code.
This new curve resulted from fatigue tests conducted on smooth specimens incorporating flush-ground
butt welds.

The level of safety incorporated in this design curve was expressed in probabilistic terms. The curve
was located 4 standard deviations of log N below the mean curve, which corresponded to a probability
of survival of better than 99.95%. The corresponding safety factors were 2,2 on stress or 15 on life.,
uniformly over the full range of application. as is the curve was linear (slope – 3,5 in log/log
coordinates).

The BS 5500 curve was lower than the ASME curve, reflecting the fact that even in the absence of a
notch, metallurgical modifications and defects inherent to welding result in a fatigue strength reduction
in the welded zones.

Furthermore, this curve was applicable to all steels, regardless of their tensile strengths. It could also be
used for aluminium alloys, by factoring the stress in accordance with the difference in elastic modulii.

FDS DOC (2001) 06 (rev.3) Annex 6 / 301

This reflects one of the important conclusions drawn from fatigue tests performed on welded assemblies
in different steels. In contrast to what is observed for unwelded materials, the fatigue strengths of
welded joints are almost independent of the tensile strength of the base metal (See Figure 2).

This property can be explained by the fact that, in welded joints, even ground flush, the fatigue life is
mainly spent in propagating a fatigue crack, as crack-like micro-defects always exist in welds. Then, on
the one hand the crack propagation rate is almost insensitive to the tensile strength of the steel, while
on the other hand it does not vary significantly from one material to the next when referred to the
modulus of elasticity.
Referring to Fig. 2, the behaviour of the notched specimen lies between the behaviour of the smooth
specimen and that of the welded specimen. This illustrates the now a well-known result that notch
sensitivity increases with the tensile strength. This sensitivity is pushed to an extreme in the case of
welded assemblies, due to the sharpness of the notches they present.

On the contrary, in unwelded materials, the time to crack initiation represents the greatest part of the
life, and this is one of the fundamental reasons why the fatigue assessment of welded joints, on the
basis of the fatigue curves of the base metal (unwelded), is incorrect and can lead to overestimates of
fatigue life.

Historically, the fatigue curve in the former Annex C of BS 5500 represented a first step towards a
correct analysis of the fatigue behaviour of welded pressure vessels. However, it addressed only a part
of the problem, namely a fatigue curve reflecting the behaviour of welded material. With regard to the
choice of stress concentration factors, Annex C brought nothing new.

This second issue was addressed in 1988 when the rules developed for fatigue analysis of welded
joints in the metallic construction in general were for the first time adapted for use in pressure vessel
design. This resulted in the publication of Enquiry Case BS 5500/79 [15]. To show the evolution,
fatigues curves of the Enquiry Case BS 5500/79, BS 5500/Appendix C (up to edition 85) and ASME
Section VIII – Division 2 are reproduced on Figure 3. A good insight on this subjects can be found in
reference [15 A].

In Germany, the same concept was adopted in the 1995 edition of AD-Merkblatt S2 [16] for fatigue
assessment of welded zones, the previous AD S2 rules being kept for the assessment of unwelded
parts.

The draft European Standard for unfired pressure vessels [2] was built on the same model, i.e. making
use of the "classical" approach for unwelded parts while switching to the new approach for welded
joints.

This new approach is detailed in next sub-section.

1.1.3) Rules developed for fatigue assessment of welded joints in the metallic construction

Since 1980, there has been a considerable development of the fatigue analysis of the welded
construction, in particular under the pressure of the needs coming from the construction of large
structures (bridges, offshore structures, etc.). Important work has been carried out at national and
international level, which has led to international rules such as the Recommendations of the IIW [17]
and the EUROCODE 3 [18]

This was due to a general consensus on the interpretation of fatigue test results and the basic principles
to be contained in fatigue rules for welded assemblies. For more details the reader can report to the
reference documents written by S. Maddox [19] or D. Radaj [20].

It has been recognised for some time that the local geometry of welds varies widely, with the result that
the stress concentrations caused by welds have a random character. This is one of the main sources of
scatter in fatigue lives. This has led to the proposal that fatigue rules for welded joints should be based
on test results obtained from actual welded components, to give a series of fatigue curves rather than a
single curve (that of the unwelded base metal).

1.1.3.1) Fatigue curves of welded assemblies

FDS DOC (2001) 06 (rev.3) Annex 6 / 302

Fatigue curves are obtained from tests on actual welded assemblies manufactured according industrial
current practice. They are expressed in terms of nominal stress range versus number of cycles. The
results include the effects of all the parameters having an influence on the fatigue resistance of the
welded assembly:

- Geometric global configuration of the welded assembly;

- Local geometry of the weld bead(s);

- Direction of the applied loading;

- Residual stresses due to welding;

- Metallurgical modifications due to welding;

- Presence of defects inherent to the welding process.

For the time being, the application field of the resulting curves is limited to common arc welding
processes. Other processes, such as electron beam, laser, etc., are not specifically covered.

Statistical evaluation shows that a regression straight line in log/log coordinates can represent the
experimental results. The security to apply to the experimental mean curve is defined in probabilistic
terms, doing the hypothesis that the results, in stress ranges ∆σ R and number of cycles N, are
distributed according a log-normal law.

Most of the rules for mechanical construction have retained a design curve two standard deviations
below the mean curve, which gives a probability of survival of 97,7 %.

Equations of the design curves are of the form:

N= C (1.1.3-1)
(∆σ R )m
Where m and C are constants.

For constant amplitude loading, the endurance limit (i.e. stress range below which the fatigue life can be
assumed to be infinite) corresponds to the stress range at 5. 106 cycles. The corresponding stress
range for variable amplitude loading is that at 108 cycles.

The value 3 given to m for N < 5. 106 cycles is somewhat conventional. It generally reflects slopes
observed experimentally. It corresponds well to the results obtained from large-scale specimens
(beams, vessels, tubular nodes), while small-scale specimens may give different slopes.

For N > 5. 106, the value of m = 5 results from the use of the linear cumulative damage rule (Miner’s
rule) for assessing the damaging effect of stresses below the constant amplitude endurance limit. Such
stresses cannot be neglected in welded structures because they gradually become damaging after
cracks have initiated under stresses above the endurance limit. It is only with this rule that the value
must be used.

This approach is best suited for regular distribution spectra (random loadings).

1.1.3.2) Classification of welded assemblies

To simplify the fatigue analysis, various weld details are combined in a limited number of classes. To
each class is associated a fatigue reference curve. Figure 4 reproduces the reference curves
associated with the classes given in EUROCODE 3.

The class is identified by a number numerically equal to the stress range at N = 2. 106 cycles expressed
in MPa.

FDS DOC (2001) 06 (rev.3) Annex 6 / 303

For each welded assembly, the class is given in a table. An assembly may have different classes,
according to the direction of the loading and the mode of cracking (See figure 5).

1.1.3.3) Stresses to consider

a) Nominal stress

The nominal stress is a reference stress, which would exist in the absence of a discontinuity. It is the
stress used in the fatigue tests, under simple loads, to characterise the level of loads using elementary
theory of structures. It ignores all stress concentrations due to discontinuities, which can make the
stress distribution different from the uniform distribution (membrane) or linear distribution (bending).

Thus, in the case of a welded attachment (See figure 6), the nominal stress ignores the stress
concentration due to the notch at the weld toe and the perturbation brought by the welded attachment in
the distribution of the longitudinal stress (a perturbation which cannot be computed by the elementary
theory of structures).

The nominal stress is that to which most of the tests and the associated classes refer.

In the fatigue analysis of structures, it has the great advantage, when it is possible to use it, to simplify
stress calculations, by avoiding the modelling of structural details.

b) Nominal stress on throat section

In the particular case of fillet welds transmitting loads, in which cracks are likely to initiate not only at the
weld toe, but also at the weld root, a particular stress is used: the nominal stress on throat section,
defined as the modulus of the mean stress vector acting on the weld throat (See figure 7).

In practice the determination of this stress is not easy, except for straight or axisymmetric assemblies
uniformly loaded on their whole length.

c) Structural stress

When the structure becomes more complex the nominal stress cannot be used. For the study of
complex geometries the structural stress is used.

The structural stress is the linearly distributed stress across the section thickness, which arises from
applied loads (forces, moments, pressure, etc.) and the corresponding reaction of the particular
structural part.

Structural stress includes the effects of gross structural discontinuities (e.g. branch connections,
cone/cylinder intersections, vessel/end junctions, thickness changes, deviations from design shape,
presence of an attachment). However it excludes the notch effect of local structural discontinuities (e.g.
weld toe), which give rise to non-linear stress distributions across the section thickness (See figure 6).
In traditional pressure vessel design terminology, it is the primary plus secondary stress.

For the purpose of a fatigue assessment, the structural stress shall be evaluated at the potential crack
initiation site, also called the hot-spot.

Structural stresses may be determined by one of the following methods:

- numerical analysis (e.g. finite element analysis with shell or plate elements),

- strain measurement and extrapolation to the hot spot or

- the application of stress concentration factors to nominal stresses obtained analytically.

Guidance on the use of numerical analysis is given in reference [21]. Guidance on the extrapolation of
strain measurements is given in clause 18 of prEN 13445-3.

FDS DOC (2001) 06 (rev.3) Annex 6 / 304

1.1.3.4) Fatigue criterion

In the case of uni-axial loading (as in most fatigue tests), the fatigue criterion is the stress range. In the
case of complex loading (bi-axial1, non-proportional), a fatigue criterion needs to be defined in order to
apply the fatigue curves obtained under uni-axial loading.

Many works have been devoted to the understanding of multi-axial fatigue, and numerous fatigue
criteria have been proposed (Sines, Crossland, Dang Van, etc.). But if these criteria reflect the
behaviour of unwelded zones, they cannot be used to represent the behaviour of welded assemblies. In
the present state of knowledge, the practical criterion that best reflects the cracking of welded zones is
the principal stress range, at least in the case of cyclic loading in which the directions of the principal
stresses do not vary.

In welded joints, which present different fatigue strengths in the longitudinal and transverse directions2,
it is generally accepted that the leading principal stress acts in the weak (transverse) direction as long
as it is not oriented at less than 30 ° (or 45 °according to the rules considered) of the strong
(longitudinal) direction.

There will be cases where it will be possible to draw benefit from a favourable orientation of the
stresses: those where the leading stress acts in the strong direction. Nevertheless, verification in the
weak direction is still necessary, under the other principal stress.

1.1.3.5) Influence of mean stress and residual stresses

The fatigue rules for welded assemblies use the full stress range regardless of the mean stress during
the cycle. This is an important difference with the rules for unwelded zones.

The reason for this difference of treatment is that welds contain high tensile residual stresses, reaching
the yield strength of the material [19].

The superposition of the applied cycle to the pre-existing stress results in an effective maximum mean
stress of:

Rp0,2 - σ max (1.1.3-2)

2

whether the cycle is in tension or in compression (See figure 93).

Therefore, the real mean stress in the material remains independent of the applied mean stress and its
effect is included in the fatigue curves for the welded assemblies.

The stress relaxation brought by Post Weld Heat Treatment (PWHT) has limited effects on the residual
stress field, because residual stresses of a level up to the yield stress of the material at PWHT
temperature may remain. So it is prudent to consider that better fatigue curves cannot be applied to
welded assemblies after PWHT.

1.1.3) The "welded joint" approach in PD 5500 (since BS 5500 ed.1988) and AD-Merkblatt S2
(since ed.1995)

Before the work on CEN harmonised standards on pressure equipment was started, two European
pressure vessel codes had already changed their fatigue rules to new ones based on the
"welded joint" approach initially developed in the metallic construction field :

− first, the British Standard BS 5500 (today PD 5500), by means of the optional enquiry case n°79, as
soon as 1988, and

1
The bi-axial loading is the situation encountered on the surface of the weld.
2
For example, a full penetration butt weld made from one side without backing.
3
In figure 9, σ e is used for the yield strength.

FDS DOC (2001) 06 (rev.3) Annex 6 / 305

− later on, the German code AD Merkblatt S2, in 1995.

In 1996, enquiry case n°79 was integrated in the main text of the British code, as Annex C, making the
corresponding rules mandatory.

In these two codes, the necessary adaptation of the "welded joint" approach to the context of pressure
vessels resulted in some differences with the original rules. The most significant are the following :

1.1.3.1 Use of structural stress/nominal stress

In beam structures, the most usual stress type used for fatigue assessment is the nominal stress,
because this stress is usually well defined at welded joints. This is not the case in pressure vessels
made with plate and shell components, due to the more complex stress pattern.
Therefore PD 5500 and AD Merkblatt S2 state that the stress which is normally to be used is the
structural stress. Use of the nominal stress is restricted to assessment of a limited number of weld
details (mainly the weld-on attachments).

1.1.3.2 Safety of the design fatigue curves and associated welded details

PD 5500 has retained design fatigue curves which are consistent with those used in the metallic
construction field, i.e. curves that are placed two standard deviations below the mean.

The German rules AD Merkblatt S2 have proposed different curves. Their number is reduced to only
four, and the associated welded details classification is such that a lower class is systematically
allocated to the details. A reason put forward to justify this increased conservatism is that a higher
safety margin is needed in pressure vessels, due to the potential high danger of a failure. This is also a
way to account for the fact that no partial safety factors on load are usually used in pressure vessel
calculations.
It is worth noticing that in prEN13445 the design fatigue curves have also been given an increased
safety, but through another solution, which is the use of curves three standard deviations below mean
(see 1.2.1.1). The resulting differences with AD Merkblatt S2 can be seen in figure 11.

Another difference lays in the number of cycles at which the fatigue curves reach the endurance limit (at
constant stress cycling): it has been placed at 2⋅106 cycles, instead of 5⋅106 in the curves used in
metallic construction. Thus the increased conservatism that affects the curves in the domain of finite
lifetime is reduced for infinite lifetime.

For both codes, it was necessary to take into account some types of welded joint that are usual in
pressure vessels (e.g. nozzle-to-shell junctions) but are not typical in metallic construction and so were
not addressed in the classification system established for that field. For such cases, the fatigue classes
allocated to the various weld details were selected by assimilation to the closer basic details listed in the
existing classification.

As regards assessment of unwelded parts, PD 5500 makes use of a unique curve derived from
experiments on unwelded structural beam profiles. As opposed to this, AD Merkblatt S2 has kept for
unwelded parts the "classical" approach already mentioned (see 1.1.1).

1.1.3.3 Extension to low cycle range

The original fatigue curves for welded details result from experiments having mainly covered the range
104 cycles to 107 cycles. Since many pressure vessels are cyclically loaded in the low cycle range, it
was necessary to extend the curves below 104 cycles, down to 100 cycles.
The solution chosen in both codes was to use a linear extrapolation (in log-log coordinates). This is
known to put the results on the safe side.

Another aspect of operating conditions in the low cycle range is the need to account for elastic-plastic
cycling. This is made by applying plasticity correction factors on the elastically calculated stresses. The
same factors are used in both codes. The studies from which they were derived are referenced in
Annex 1 to AD Merkblatt S2.

FDS DOC (2001) 06 (rev.3) Annex 6 / 306

1.2) The European standard for unfired pressure vessels (prEN 13445-3)

This clause is devoted to a presentation of the detailed and simplified fatigue analysis, as provided
respectively in clause 18 and 17 of the draft European Standard on Unfired pressure vessels.

1.2.1) Detailed fatigue analysis (Clause 18)

The rules are applicable for fatigue assessment of pressure vessel components, which are subjected to
repeated fluctuations of stress. The assessment procedure assumes that the vessel has been designed
in accordance with all other requirements of the standard (e.g. design rules, manufacturing
requirements and non destructive testing of welded joints).

These rules are only applicable to ferritic and austenitic steels according to prEN 1345-2. Fatigue
design curves can be used up to 380 °C for ferritic steels and 500 °C for austenitic steels. The fatigue
strengths specified do not include any allowance for corrosion.

These rules are not applicable to vessels subjected to visual inspection only vessels (testing group 4
vessels, see further). For vessels subjected to random testing (testing group 3 vessels, see further),
special provisions apply. This method is not intended for special design involving elastic follow-up [22].

The fatigue evaluation differs whether it is a welded or an unwelded zone.

1.2.1.1) Welded zones

The fatigue design rules for welded zones are similar in principle to those developed for metallic
construction. However, they include some important differences due to the specific characteristics of
pressure vessels:

- Structures made of shells, with significant curvature;

- Subjected to biaxial loading, mainly due to pressure;

- Thermal stresses may play an important role.

a) Design fatigue curves

The design fatigue curves are consistent with those of EUROCODE 3 [18]. They are linear in log-log
coordinates and conform to the equation:

N= C (1.2.1-1)
(∆σ R )3
As earlier, the design curves are identified by a number called “Class” numerically equal to the fatigue
strength value ∆σ R (N/mm²) at a fatigue life N = 2. 106 cycles. There are ten curves corresponding to
classes: 100, 90, 80, 71, 63, 56, 50, 45, 40 and 32 (See figure 10). Each curve applies to particular
welded assemblies, called “construction details” in the Standard. The endurance limit under constant
cyclic load is reached at 5.⋅106 cycles.

For lives above 2. 106, the curves, which are consistent with those of EUROCODE 3, may be
conservative. Alternative curves and constant amplitude endurance limits are permissible if they can be
justified by the user.

In AD-Merkblatt S2, 1995 edition [24], the endurance limit is reached at 2. 106 cycles. However, it may
also be noted that the endurance limit in PD 5500 corresponds to 5.106 cycles, reflecting the fact that
there are fatigue data for low Class details that lie below the Eurocode 3 endurance limit

As in PD 5500 and AD Merkblatt S2, a major difference with the curves of EUROCODE 3, which are
limited to lives exceeding 104 cycles, is their extension down to 100 cycles to cover low-cycle fatigue,

FDS DOC (2001) 06 (rev.3) Annex 6 / 307

which very often affects pressure vessels, in particular the full pressure cycles between 0 and the
maximum allowable pressure Ps . This extension is carried out by simple extrapolation of the curves,
which are linear in log/log coordinates.

This extrapolation is certainly conservative and justified by available low-fatigue data for welded joints.
Comparisons with low-cycle fatigue test results obtained from welded pressure vessels, have been
presented by the authors of the Enquiry case BS 5500/79 to support the validity of this extrapolation
[23].

The curves have been derived from fatigue test data obtained from appropriate laboratory specimens,
tested to rupture under load control, or for applied strains exceeding yield (low cycle fatigue), under
strain control. Continuity from the low to high cycle regime is achieved by expressing the low cycle
fatigue data in terms of the pseudo-elastic stress range (i.e. strain range multiplied by the elastic
modulus, if necessary corrected for plasticity).

The fatigue strength design curves are approximately three standard deviations of log (N) below the
mean curve, fitted to the original test data. Thus, they represent a probability of failure of approximately
0,1 %. This probability of failure is much lower than that given by EUROCODE 3: 2,3%, but higher than
that given by AD-Merkblatt S2, 1995 edition: 0,01 %.

Figure 11 (already mentioned in 1.1.3.2) reproduced from a paper by Bernhard Gorziske [25] shows the
differences between the fatigue curves of the European Standard, EUROCODE 3 and AD Merkblatt S2,
1995 edition, for the same weld construction detail: a full penetration butt weld made from both sides.
Considerable differences appear in the classes, which are respectively: 80, 90 and 63.

Annex C of the British Code PD 5500:2000 [26] has design fatigue curves consistent with those of
EUROCODE 3.

A European consensus on the safety coefficients of the fatigue curves would be very important for the
future.

b) Stresses to consider

If the potential fatigue crack initiation site is at the weld toe or on the surface of the weld, the structural
stress range shall be evaluated at the potential crack initiation site.

In most cases, the structural stress is the linearly distributed stress across the section thickness, as for
metallic construction. However, under high thermal stresses or with mechanical loads developed in thick
components, the linearly distributed stress is inadequate. Then the structural stress is the stress
determined by taking into account the real geometry, but excluding minor discontinuities sources of
peak stresses.

The best way to compute the structural stress is to extrapolate stresses acting some distance from the
point under consideration. Stresses to be used are:

- Either total stresses resulting from a refined numerical analysis with solid finite elements through
the thickness,

- Or stresses measured experimentally on the surface with strain gauges in the vicinity of the point
under consideration.

The influence zone of a minor discontinuity being of the order of 0,4 times the thickness of the wall, the
extrapolation shall be done in considering only the stresses exerted beyond that distance. The
extrapolation is linear or quadratic and extended to all the components of the stress tensor (See figure
12).

As the stress state is three-dimensional, an equivalent structural stress range ∆σ eq is determined by

application of the Tresca criterion. A detailed procedure is provided in the standard for this calculation.
∆σ eq is a scalar and has no direction.

FDS DOC (2001) 06 (rev.3) Annex 6 / 308

Clause 18 also offers the possibility to use the structural principal stress range. Since one principal
stress normal to the surface is equal to 0 and can be neglected, use is made of the two other structural
principal stresses acting essentially (i.e. within 45 °) parallel and normal to the direction of the weld
respectively.

For welded attachments the nominal stress can be used.

In case of directly loaded fillet welds or partial penetration welds, the maximum range of stress on the
weld throat is used and a unique class 32 is applied. Nevertheless this type of analysis may be avoided
if the weld throat thickness is sufficient.

Directly loaded fillet and partial penetration welds do not need to be assessed if the effective weld throat
thickness is such that the stress range in the weld does not exceed 0,8 times the stress range in the
plate. Welds not subject to direct loading (e.g. attachments) do not need to be assessed if the effective
weld throat is at least 0,7 times the thickness of the thinner part joined by the weld.

c) Classification of weld details

Classification of weld details has been established on the basis of nominal stresses acting in one
direction, normal or parallel to the weld. Therefore different cases shall be considered for use of the
design fatigue curves with the structural stress.

When the equivalent structural stress range is used, the class indicated in table 18-4 of the standard
refers to the least favourable direction for the particular weld detail and mode of fatigue cracking shown,
since ∆σ eq has no direction.

When the principal stress range is used, advantage may be taken of the direction of the loading. Two
classes are given for a weld detail in table 18-5. If the maximum principal stress range acts in the
direction, which is within 45° of the direction of the weld, then the weld can be classified as being
parallel to the direction of the loading with respect to the maximum principal stress range and normal to
the direction of the loading with respect to the minimum principal stress range.

Unless expressly mentioned, the fatigue design curves do not incorporate the effect of deviations from
the ideal shape, such as misalignment or peaking. These have the effect of introducing secondary
bending stress when the welded joint is loaded and therefore their effect must be included in the
determination of the stresses.
In the metallic construction codes, the situation is different because the fatigue curves for butt joints do
cover a limited amount of shape deviation (bounds are given). Thus the fatigue classes for butt joints
details are lower.

The fatigue class may be downgraded if the Non Destructive Testing (NDT) performed on the weld is
not sufficient to confirm the absence of significant flaws. A special column is reserved in tables 18-4 and
18-5 for random testing (column named “testing group 3”, see also 1.2.3 further).

d) Correction factors for welded zones

Fatigue curves of welded details are valid for:

T ≤ 100 °C

e ≤ 25 mm

Thus, corrections for temperature and thickness are applied when these limits are exceeded.

Tests [27] have shown that a reference temperature T * less than the maximum cycle temperature
could be used. This temperature is given by:

T * = 0,75⋅Tmax + 0,25⋅Tmin (1.2.1-2)

FDS DOC (2001) 06 (rev.3) Annex 6 / 309

The temperature correction factor for ferritic material is:

fT * = 1,03−1,5⋅(10) ⋅T *−1,5⋅(10) ⋅T *²
−4 −6
(1.2.1-3)

and for austenitic material:

fT * = 1,043−4,3⋅(10) ⋅T *
−4
(1.2.1-4)

The thickness correction factor is:

( )
few = 25
en
0,25
(1.2.1-5)
The overall correction factor is:

fw = fT *⋅ few < 1 (1.2.1-6)

And the number of allowable cycles is:

N= C (1.2.1-7)
m
 ∆σ eq 
 fw 
 
since no correction for mean stress is necessary.

The fatigue damage D corresponding to the application of n cycles is:

D= n (1.2.1-8)
N

1.2.1.2) Unwelded zones

The method of fatigue analysis of unwelded material is based on that in AD-Merkblatt S 2 [24].

a) Design fatigue curves

The design curves have been derived from fatigue test data obtained from un-notched polished ferritic
and austenitic rolled and forged steel specimens at room temperature, under alternating (mean load =
0) load control or, for applied strains exceeding yield (low-cycle fatigue), strain control. The failure mode
is crack initiation.

Compared to the mean curve fitted to the original data, the curves incorporate safety factors of 10 on
fatigue life and 1,5 on stress range. Design fatigue curves have the form

 4,6.(10 ) 
4
N =  (1.2.1-9)
 ∆σ R - 0,63⋅ Rm + 11,5 

They are represented in figure 13.

b) Fatigue parameter

The fatigue parameter is the notch or total stress that includes the effect of all the discontinuities (gross
and local discontinuities).

To be consistent with the analysis of welded zones, equivalent effective notch stresses are determined
using structural principal stresses, which incorporate the full effect of gross and local structural
discontinuities. The equivalent effective notch stress can be obtained by the use of Keff given by the
following empirical formula [28]:

FDS DOC (2001) 06 (rev.3) Annex 6 / 310

 1,5⋅(K t −1)
Keff = 1 + (1.2.1-10)
 K ⋅∆σ struc 
1 + 0,5⋅max1; t
 ∆σ D 

where ∆σ D =∆σ R for N ≥ 2⋅(10) cycles.

6

c) Correction factors

Fatigue analysis of unwelded zones includes correction factors for:

- Temperature fT *

- Thickness fe

- Surface finish fs

- Mean stress fm

which are taken from AD-Merkblatt S2.

Therefore the number of allowable cycles at a specific stress range ∆σ eq is given by:

 
4,6.(10)
4
 
N = (1.2.1-11)
∆σ eq - 0,63⋅Rm + 11,5 
 fu 
 

With:

fT* ⋅ fe⋅ fs⋅ fm

fu = (1.2.1-12)
Keff

The fatigue damage D corresponding to the application of n cycles is:

D= n (1.2.1-13)
N

d) Simplified procedure

A simplified procedure for the fatigue assessment of unwelded steel is permissible using class 100
design data for welded components, independently of material static strength or surface finish. The data
are used in conjunction with equation 1.2.1-7, with fw replaced by f u* :

fe⋅ fT*⋅ fm
fu* = (1.2.1-14)
Keff

1.2.1.3) Common rules

The common rules deal with plasticity correction and the definition of stress cycles when considering
variable amplitude loading.

a) Plasticity correction

FDS DOC (2001) 06 (rev.3) Annex 6 / 311

For both welded and unwelded zones, plasticity corrections need to be applied if the calculated pseudo-
elastic structural stress range exceeds twice the yield strength of the material under consideration.

Correction factors proposed for mechanical loading Ke and thermal loading Kν are issued from studies
performed in Germany and France.

- For mechanical loads, results of studies performed in Germany [29] are represented in figure 14.
Formulae provided in prEN 13445 are based on theses studies. They are simpler to use than
those given in AD-Merkblatt S2 and give almost the same results.

A basic objection is that Ke is normally applied to the total stress and not to the structural stress,
if one refers to the studies on which this factor has been based. However the conservatism of
the fatigue curves for welded joints in the low-cycle regime is such that this approximation
appeared valid to the experts of the Standardization Committee. This opinion is supported by
Bernhard Gorsitzke [25]. It has also been considered acceptable for application to unwelded
parts.

- For thermal loadings with a non-linear distribution through the material thickness, the correction
is based on the studies performed by J.M. Grandemange and co-workers [30]. Again the
correction is applied to the structural stress and not to the total stress.

- For combined loadings, a separate correction is applied to mechanical and thermal loadings.

b) Definition of stress cycles

For determining the stress cycles very simple rules are proposed in prEN 13445.

- A simplified cycle counting method, when the loading sequence is unknown;

- The reservoir cycle counting method, when the loading sequence is known.

These rules, which assume the simple hypothesis of synchronous variations of the components of the
stress tensor, do not reflect the complexity of some actual loadings.

1.2.1.4) Fatigue strength of steel bolts

The standard contains rules for axially loaded steel bolts, based on those in BS 7608:1993. The fatigue
strength is expressed in terms of the ratio:

maximum nominal stress range

= ∆σ (1.2.1-15)
nominal ultimate strength of bolt material Rm

The single design curve:

N = 285 3
( )
(1.2.1-16)
∆σ R
Rm

with an endurance limit at 2. 106 cycles, shown in figure 15, is used for any thread form (machined,
ground or rolled) and core diameters up to 25 mm. However, regardless of the actual tensile strength of
the bolt material, Rm should never be assumed to exceed 785 MPa.

1.2.2) Simplified fatigue analysis

Clause 17 of prEN 13445-3 deals with the simplified fatigue analysis. Pressure vessels concerned are
those for which Design by Formulae according to Clauses 7 to 16 has been applied. This analysis is not
applicable for vessels subject to visual inspection only (testing group 4 vessels, see 1.2.3 further).

FDS DOC (2001) 06 (rev.3) Annex 6 / 312

Fatigue design curves for welded details are identical to those given in Clause 18. The unique curve for
unwelded zones is a lower bound that takes into account surface roughness up to that of rolled or
extruded surfaces and also covers the effect of maximum possible mean or residual stresses.

The purpose of this method is to avoid a detailed stress analysis by finite elements or analytical
methods. Therefore the stress range due to a pressure variation ∆P , in a critical zone of a component,
is estimated by a simple formula:

∆σ = ∆P ⋅η⋅ f (1.2.2-1)
Pmax
where:

η is the stress factor for a component given in table 17-1.

f is the nominal design stress.
Pmax is the maximum permissible pressure of the component, based on the actual thickness.

For simplification the maximum permissible pressure of the whole vessel may be used instead of that of
the component. It is also possible, but conservative to use the calculation pressure P of the component.

If they are known, stress ranges (calculated as structural stresses) resulting from loads other than
pressure cycles can be taken into account in the simplified analysis by:

- adding these stress ranges to the stress range due to pressure according to equation (1.2.2-1) if
these load cycles act simultaneously to the pressure range or

- adding the fatigue damage resulting from these stress ranges to the damage resulting from the
pressure cycles according to equation (1.2.2-1) if these load cycles do not occur simultaneously.

Fatigue design curves and corrections are similar to those given for detailed fatigue analysis.

1.2.3) Relation of fatigue analysis with the rest of the standard

CEN experts have given special attention to the coherency of the fatigue rules with:

- The type of analysis performed;

- Non Destructive Testing (NDT) of the welds;

- The manufacturing tolerances;

- In-service inspection.

Links between fatigue analysis and manufacturing quality and NDT are summarised in figure 16.

1.2.3.1) The type of analysis

Part 3 of the standard provides rules for the design of pressure vessels or pressure vessel components
using design by formula (DBF).

In addition, two other methods can be used to supplement or replace DBF:

- Design by Analysis (DBA), the rules of which are given in annexes B and C;

- Experimental techniques.

Experimental techniques are not covered in the first edition of the standard.

By limiting the stresses to elastic shakedown, the requirements specified in clauses 6 to 16 (DBF) will
provide satisfactory designs for pressure loading of predominantly non-cyclic nature, i.e. when the

FDS DOC (2001) 06 (rev.3) Annex 6 / 313

number of full pressure cycles (0 to Ps ) or equivalent full pressure cycles is less or equal to 500. For
pressure cycles at a pressure ∆Pi less than the full pressure, the number of equivalent full cycles n eq is
given by:

3
 ∆Pi 
neq = ni   (1.2.3-1)
 Pmax 

where Pmax is the maximum permissible pressure obtained with the analysis thickness4 (minimum
possible thickness minus the corrosion allowance) and the calculation temperature.

Therefore the number Ni of allowable pressure cycles of range ∆Pi is:

3
P 
Ni = 500  max  (1.2.3-2)
 ∆Pi 

Exponent 3 of the formulae comes from fatigue design curves for welded components.

If the number of full pressure cycles or equivalent full pressure cycles is likely to exceed 500, the
calculations shall be completed by a simplified fatigue analysis, as given in clause 17 or, if necessary,
by a detailed fatigue analysis, as given in clause 18.

Pressure vessels subjected to visual testing only, under testing group 4, as defined in EN 13445-5, are
intended for predominantly non-cyclic operation and are limited to 500 full pressure cycles or equivalent.

If a component is subject to loading other than pressure, or if for a component no requirements are
supplied in clauses 6 to 16, the rules given in annexes B and C (Design by Analysis) shall be applied
and the fatigue analysis performed according to clause 18.

Fatigue analysis may be avoided by reference to previous industrial experience.

1.2.3.2) Non destructive testing

The fatigue resistance of welds can be reduced as a result of the presence of weld imperfections.
Therefore a link has been established between the fatigue analysis and the NDT during manufacturing,
as specified in Part 5.

Each vessel is normally allocated to a single testing group5,6. There are the four testing groups designed
to reflect the extent of non destructive testing and destructive testing in association with weld joint
coefficient, material grouping, welding process, maximum thickness, service temperature range.
Briefly:

- Testing group 1 means 100 % volumetric testing and 10 % surface testing of the governing
joints7. Then the joint factor of governing joints is equal to 1. The extent of testing of other welds
is as specified in Part 5;

- Testing group 2 means (100-10 %) volumetric testing and 10 % surface testing of the governing
joints. In the brackets, the first figure applies initially, the second after successful experience.
Then the joint factor is equal to 1. Extent of testing of other welds is as specified in Part 5.
Application of testing group 2 requires special conditions, in particular fully mechanised welding
and specific material groups;

4
The analysis thickness is greater than the required thickness.
5
The testing group of a vessel is not linked to the hazard category.
6
Mixing of testing groups is allowed under certain conditions.
7
Longitudinal welds in cylinders and cones, welds in spheres and dished ends

FDS DOC (2001) 06 (rev.3) Annex 6 / 314

 - Testing group 3 means (25 or 10 %) volumetric testing and 10 % surface testing of the
governing joints. Then the joint factor is equal to 0,85. Extent of testing of other welds is as
specified in Part 5;

- Testing group 4 means visual testing only. Then the joint factor is equal to 0,7. Severe limitations
apply to service conditions and materials. A higher pressure test shall be applied. (See Part 5).

For pressure loading of predominantly non-cyclic nature, as defined earlier, acceptance criteria for weld
imperfections are those referred as Quality Level C in the international standard ISO 5817:2000 [31] for
visual testing and by the corresponding standards for each testing method. For more details, see Part 5.

When the number of equivalent full pressure cycles exceeds 500, a fatigue analysis shall be performed
according to clause 17 or 18 and a more stringent inspection of welds shall be necessary:

- No root concavity;

- No lack of penetration for full penetration welds, “Level C” acceptance criteria apply to partial
penetration welds only;

- No undercut;

- Welds with smooth transition;

- NDT, as specified in Part 5, of all critical points. Then more stringent “Level B” acceptance
criteria shall apply.

According to the definition given by WG C of TC 54, critical points are those for which the limit value
Dmax of the damage factor is:

- Dmax = 0,8 for 500 < neq ≤ 1000

- D max = 0,5 for 1000 < neq ≤ 10000

- D max = 0,3 for neq > 10000

1.2.3.3) Manufacturing tolerances

Tests of welded specimens are performed on perfect geometries and deviations from the ideal
geometry like misalignment are not included in fatigue design curves. As noted earlier, such deviations
introduce secondary bending and so increase the stress experienced by the welded joint. When, in rare
cases, deviations from shape are taken into account, they are explicitly mentioned.

It is therefore the responsibility of the designer, in co-operation with the manufacturer, to assume certain
manufacturing tolerances at the design stage and to increase the stress range to be used with the
fatigue curve accordingly. These tolerances must not be exceeded during the manufacture of the
vessel. Otherwise it is necessary to repeat the fatigue assessment.

Tolerances given in Part 4 for components subjected to predominantly non-cyclic loading (less than 500
full pressure cycles) may be inappropriate for cyclic loading.

Clause 18 provides guidance on the calculation of the increases in stress due to the following deviations
from perfect shape:

- Axial misalignment;
- Ovality;
- Angular distortion.

acting alone or simultaneously (See figure 17).

FDS DOC (2001) 06 (rev.3) Annex 6 / 315

1.2.3.4) In service inspection

In clause 18.4.4 reference is made to in-service inspection. It is written:

“It is a condition of the use of these rules that all regions which are fatigue-critical are accessible for
inspection and non-destructive testing, and that in-service inspection shall be preformed at no later than
20 % of the allowed fatigue life”.

These requirements are completed by an annex M that gives recommendations for measures to be
adopted in service:

The time of the first inspection is connected to the fatigue design curves. As a result, in-service
inspection is only required after 50 % of the design fatigue life, as in AD-Merkblatt S2 [24].

1.3) The European standard for metallic industrial piping

The draft European standard for metallic industrial piping prEN 1480 deals with fatigue analysis in Part
3.

1.3.1) Detailed analysis

Detailed analysis is required in clause 12.4 if the conditions of clause 10 are not met. However no rules
are provided for this analysis. Reference is made to Part 3 of the water tube boiler standard prEN 12952
or to Part 3 of the unfired pressure vessel standard prEN 13445.

“The method of calculation in prEN 12952-3, including appropriate annexes, may be used to provide an
acceptable analysis where external forces and moments are not significant.

Alternatively the method of calculation in prEN UFPV-3 may be used”.

1.3.2) Simplified analysis

The simplified analysis of piping, given in Clause 10, is limited to pressure loads. This analysis is in
conformity with AD-Merkblatt S1 [32].

1.4) The European standards for boilers

1.4.1) Water tube boilers

The fatigue analysis contained in Annex B of prEN 12952-3 is a detailed analysis using fatigue design
curves of smooth specimens. The fatigue analysis is similar to that contained in AD-Merkblatt, 1988
edition, [11]. The approach belongs to the classical approach described in Clause 1.1.1 of this report
and deserves the same comments.

1.4.2) Shell boilers

No fatigue analysis is provided in the shell boiler standard prEN

FDS DOC (2001) 06 (rev.3) Annex 6 / 316

2) Recommendations in view of improving the rules of European Standards
This second Part deals with recommendations to improve the fatigue analysis of European Standards.
They are based on the information related to code and standards obtained from:

- The results of the enquiry

- The survey of pressure vessel fatigue failure
- The survey of current R & D in support of pressure equipment fatigue design
- The survey of priority R&D requirements for pressure equipment fatigue design from industry
perspective
- The assembly of fatigue data available to facilitate pressure equipment fatigue design
- The first part of this report

2.1) General recommendations

These recommendations apply to all the standards for pressure equipment.

2.1.1) Recognition of good experience

It is important that all standards for pressure equipment contain a clause recognising that fatigue
analysis is not necessary where experience from equivalent designs and operating conditions shows
that fatigue is not a problem. The review of standards has shown that this important sentence is not yet
included in the Standard for unfired pressure vessels.

2.1.2) Promotion of the European approach to the fatigue behaviour of welded assemblies

The use of fatigue design curves of welded assemblies is not yet generalised. The water tube boiler
Standard adopts the “classical approach”. It would be highly desirable to have the same approach for all
the pressure equipment Standards. This would be a simplification for the research, which could
concentrate on well-defined objectives, such as:

- Additional research on endurance limits

- Additional research on low-cycle fatigue research and plasticity

- Search of a European consensus on safety coefficients (Safety coefficients of fatigue design

curves are different in prEN 13445, AD-Merkblatt S2 and PD 5500:2000)

2.1.3) Clarification and simplification of the technical requirements

Present rules seem complex and difficult to apply to Industry.

- The internal organisation of each standard should be reviewed in view of a better understanding
of the rules

- The number of variants should be reduced, for example the simplified fatigue analysis of
unwelded zones can be made by:

o A simplified procedure (Clause 18 of the Standard for unfired pressure vessels prEN
13445-3)

o The procedure of Clause 17 of prEN 13445-3

o A procedure similar to that of AD-Merkblatt S1 (Clause 10 of the Standard for

metallic industrial piping)

2.1.4) Reliability of fatigue analysis

The fatigue analysis proposed by European Standards is relatively new for Industry. Reliability of the
analysis is a crucial point. This reliability should be demonstrated by selected tests on components, with

FDS DOC (2001) 06 (rev.3) Annex 6 / 317

large diffusion of the results. The importance of manufacturing tolerances in the analysis should be
emphasised.

Criteria based on a fixed number of cycles (500 for the pressure vessel Standard or 1000 for the
metallic industrial piping Standard) are not always safe. It is better to perform a simplified fatigue
analysis and still better a detailed fatigue analysis, which is necessarily costly with a specialised
software.

2.1.4) Publication of criteria and worked examples

It is essential to publish criteria concerning the origin of the rules, because the present fatigue rules
differ notably from those of the ASME Code, which are familiar to Industry. Publication of criteria should
be accompanied by worked examples and if necessary by specialised software.

2.1.5) Inclusion of a clause on experimental analysis

It is important to include a Clause on experimental analysis in each Standard. A particular appraisal

may solve difficult design problems (e.g. the design of bellows) or reduce the conservatism involved in
general rules (fatigue design curves of weld details are based on a standard quality).

2.1.6) Extension of the scope to cover new items

Suggested new items are:

- Extension to new materials: aluminium and nickel alloys, etc. (Nodular cast iron is already
covered in Annex A of prEN 13445-6)

- Extension to random loading (See for instance the standard in preparation on “Gas-loaded
accumulators for fluid power applications”)

- Extension to the creep range

- Extension to corrosive environments (e.g. hydrogen or ammonia storage)

2.1) Specific recommendations

These recommendations aim to increase the use the Standards.

2.1.1) Recommendations for the unfired pressure vessel Standard

In Clause 18 of prEN 13445-3, fatigue analysis of the welded zones can be done either by use of the
equivalent stress range or by use of the principal stress range. It is suggested to clearly separate the
two analyses, by putting the analysis based on the principal stress range in a separate annex. In the
vast majority of cases Industry will use the simpler method, based on the equivalent stress range.

In the same manner it is suggested to put the simplified procedure for unwelded zones in a separate
annex to avoid confusion.

2.2) The European standard for metallic industrial piping

It does not seem advisable to recommend a detailed analysis based on the option of using either the
boiler Standard or the pressure Standard. One of them should be selected and its compatibility with the
simplified analysis verified.

2.3) The European standard for boilers

The use of fatigue design curves of welded details should be recommended.

FDS DOC (2001) 06 (rev.3) Annex 6 / 318

Conclusion
This report has shown the slow evolution that has led to the adoption of a new approach for the fatigue
design of welded zones. This analysis is based on tests performed on welded specimens in conjunction
with the use of a specific stress called structural stress. This analysis is largely used in the domain of
metallic construction, but its penetration in the pressure equipment Industry is difficult, because, apart
from the users of BS PD 5500, this Industry is more accustomed to the classical approach of the ASME
Code.

Therefore measures proposed in the second part of this report have a great importance for developing
the acceptability of the new European Standards. They aim at easing the accessibility to the Standards
and developing the reliability of fatigue design analysis.

FDS DOC (2001) 06 (rev.3) Annex 6 / 319

REFERENCES:

[1] Directive 97/23/EC of the European Parliament and of the Council of 29th May 1997, On the
approximation of the laws of the Member States concerning pressure equipment, OJEC No
L181/1-55, 9th July 1997.

[2] prEN 13445, Part 1 to 6, Unfired pressure vessels, Part 1 draft of May 2001, Part 2 draft of July
2000, Part 3 draft of July 1999, Part 4 draft of January 2001, Part 5 draft of April 2001, Part 6
draft of June 2001.

[3] prEN 12952, Part 3, Water tube boilers, draft of April 2001.

[4] prEN 12953, Part 3, Shell boilers, draft of 2001.

[5] prEN 13480, Part 3, Metallic industrial piping, draft of April 1999.

[6] BS 5500, British standard specification for unfired fusion welded pressure vessels, 1992 edition,
British Standards Institution, London.

[7] Handtschoewercker, Alain, Evolutions récentes et en cours des règles d’analyse de la fatigue
dans les Codes d’Appareils à Pression, CETIM report 194454, presented at the AFIAP
Conference, Paris, 1995.

[8] ASME Boiler and Pressure Vessel Code, Section III, Nuclear power plant components, Chapter
NB 3200, The American Society of Mechanical Engineers, New York, 1995

[9] ASME Boiler and Pressure Vessel Code, Section VIII, Division 2, Pressure Vessel – Alternative
rules, Appendix 5: Design based on fatigue analysis, The American Society of Mechanical
Engineers, New York, 1995

[10] RCC-M, Design and Construction Rules for Mechanical components of PWR Nuclear Islands,
2000 edition, AFCEN, Paris La Défense.

[11] AD-Merkblatt S2, Design to allow for fluctuating stress, English translation, 1988 edition, Carl
Heymanns Verlag KG, Köln.

[12] Criteria of the ASME Boiler and Pressure Vessel Code for Design by Analysis in Sections III
and VIII, Division 2, The American Society of Mechanical Engineers, New York, 1968.

[13] Peterson, R.E., Stress concentration factors, J. Wiley and Sons, New York, 1974.

[14] Heywood, R.B., Designing against fatigue, Chapman and Hall, 1962.

[15] Enquiry Case BS 5500/79 to BS 5500 – May 1998, Assessment of vessels subject to fatigue-
Alternative approach to method in Appendix C, British Standard Institution, London.

[15 A] Maddox, S.J., Fatigue aspects of pressure vessel design, in “Pressure Vessel Design –
Concepts and Principles”, editors J. Spence and A.S. Tooth, E&F.N. Spoon, London, 1994.

[16] AD-Merkblatt S2, Fatigue Analysis, English translation, 1995 edition, Carl Heymanns Verlag
KG, Köln.

[17] Hobbacher, A. Fatigue design of welded joints and components IIW, Abington publishing,
Abington, Cambridge, 1996

[18] EUROCODE 3, Design of steel structures - Part 1: General rules and rules for buildings, Draft
version 3, April 1990, European Community Commission.

[19] Maddox, S.J., Fatigue strength of welded structures, Abington Publishing, Abington,
Cambridge, 2nd edition, 1991.

FDS DOC (2001) 06 (rev.3) Annex 6 / 320

[20] Radaj, D., Design and analysis of fatigue resistant welded structures, Abington Publications,
1990.

[21] Niemi, E., Stress determination for fatigue analysis of welded components, IIW , Abington
Publishing, Abington, cambridge, 1995.

[22] Baylac, G., Fatigue and Inelastic analysis, in Computational Mechanics ’88, S.N. Atluri and G.
Yagawa editors, Springer Verlag, 1988.

[23] Harrison, J.D., Maddox, S.J., A critical examination of the rules for the design of pressure
vessels subject to fatigue loading, Proceedings of the 4th International Conference on Pressure
Vessel Technology, IMechE, 1980.

[24] AD-Merkblatt S2, Design to allow for fluctuating stress, English translation, Edition 1995, Carl
Heymanns Verlag KG, Köln.

[25] Gorsitzke, B., Neuere Berechnungsvorschriften zum Ermüdungsfestikeitsnachweis von

Druckbehältern, Teil 1 und 2, TÜ Bd. 36, Nr. 7/8, 1995.

[26] PD 5500, Specification for unfired fusion welded pressure vessels, Annex C, Assessment of
vessels subject to fatigue, 2000 edition, British Standards Institution, London.

[27] Wellinger, K., and Idler, R., Der Einfluß wechselnder Temperaturen auf das
Zeitfestigkeitsverhalten von Stählen, Archiv Eisenhüttenw. 48, 1977, Issue 6, pp. 347/52.

[28] Kiesewetter, N., Discussion paper to CEN/TC 54/WG C/S-G Design Criteria, 2001.

[29] Kühnapfel, K.-F., and Troost, A., Näherungslösungen zur rechnerischen Ermittlung von
Kerbspannungen bei elasto-plastischer Beanspruchung, Z. Konstruktion 31, 1979.

[30] Grandemange, J.M., et al, Improvements in Fatigue Analysis Methods for the Design of Nuclear
Components Subjected to the French RCC-M Code, WRC Bulletin.

[31] ISO/DIS 5817:2000, Welding – Fusion-welded joints in steel, nickel, titanium and their alloys
(beam welding excluded) – Quality levels for imperfections

[32] AD-Merkblatt S1, Simplified analysis for cyclic loading, May 1998 edition, Carl Heymanns

FDS DOC (2001) 06 (rev.3) Annex 6 / 321

FDS DOC (2001) 06 (rev.3) Annex 6 / 322
FDS DOC (2001) 06 (rev.3) Annex 6 / 323
FDS DOC (2001) 06 (rev.3) Annex 6 / 324
FDS DOC (2001) 06 (rev.3) Annex 6 / 325
FDS DOC (2001) 06 (rev.3) Annex 6 / 326
FDS DOC (2001) 06 (rev.3) Annex 6 / 327
FDS DOC (2001) 06 (rev.3) Annex 6 / 328
 10000

E=2,09x105 N/mm2

1000
(N/mm²)

32 40 45 50 56 63 71 80 90 100
2
100

1
10
1,0E+02 1,0E+03 1,0E+04 1,0E+05 1,0E+06 1,0E+07 1,0E+08

N
Fi
gure 10
Fatigue design curves for welded components

Class is the range at 2 106 cycles. It varies from 100 to 32 MPa

1 Curves for assessing variable amplitude loading
2 For constant amplitude loading endurance limit: range at 5 106 cycles

FDS DOC (2001) 06 (rev.3) Annex 6 / 329

 FATIGUE CURVES BUTT WELD FROM BOTH SIDES
(Constant amplitude)

10000
Stress range (MPa)

1000
Eurocode 3
prEN 13445
100 AD-S2 (1995)

10
1,0E+0 1,0E+0 1,0E+0 1,0E+0 1,0E+0 1,0E+0 1,0E+0
2 3 4 5 6 7 8
Number of cycles

Figure 11
Design fatigue curves
2 standard deviations below mean curve (EUROCODE 3 and BS 5500)
3 standard deviations below mean curve (prEN 13445-3)
4 standard deviations below mean curve (AD-Merkblatt S2)

Figure 12

Full penetration weld

1 Nominal stress far from discontinuity
2 Structural stress
3 Notch stress
4 Extrapolation to give structural stress at potential crack initiation site

FDS DOC (2001) 06 (rev.3) Annex 6 / 330

 10 000
Stress range, ∆σ R ,N/mm2

Rm (N/mm2)
1000
800
600
1 000 400

100
1,0E+02 1,0E+03 1,0E+04 1,0E+05 1,0E+06 1,0E+07 1,0E+08
Fatigue life, N, cycles

Figure 13
Fatigue design curves for unwelded ferritic and austenitic forged
and rolled steels (mean stress = 0)

FDS DOC (2001) 06 (rev.3) Annex 6 / 331

FDS DOC (2001) 06 (rev.3) Annex 6 / 332
 Figure 15
Fatigue design curve for bolts

FDS DOC (2001) 06 (rev.3) Annex 6 / 333

 Figure 16
Fatigue analysis and NDT

FDS DOC (2001) 06 (rev.3) Annex 6 / 334

 Figure 17
Deviations from design shape at seam welds

Figure 18
Weld toe dressing
Weld toe is a potential site of crack initiation
due weld shape and possible weld defects.
Grinding improves the classification of this weld detail.

FDS DOC (2001) 06 (rev.3) Annex 6 / 335