See discussions, stats, and author profiles for this publication at: https://www.researchgate.

net/publication/270917830

Comparison between the Eurocode for Fatigue of Steel Structures, EN 1993-1-

9, and the Eurocode for Fatigue of Aluminium Structures, EN 1999-1-3

Article  in  Procedia Engineering · December 2013

DOI: 10.1016/j.proeng.2013.12.060

CITATION READS

1 819

3 authors, including:

Johan Maljaars Mladen Lukic

TNO Centre Technique De La Construction Metallique
91 PUBLICATIONS   316 CITATIONS    15 PUBLICATIONS   183 CITATIONS   

SEE PROFILE SEE PROFILE

Some of the authors of this publication are also working on these related projects:

Joints in aluminium frame structures View project

International Symposium on Probabilistic Fatigue & Fracture Approaches Applied to Materials and Structures (ISPFA2017) View project

All content following this page was uploaded by Mladen Lukic on 02 October 2015.

The user has requested enhancement of the downloaded file.

 Available online at www.sciencedirect.com

ScienceDirect
Procedia Engineering 66 (2013) 34 – 48

5th Fatigue Design Conference, Fatigue Design 2013

Comparison between the Eurocode for fatigue of steel structures,
EN 1993-1-9, and the Eurocode for fatigue of aluminium structures,
EN 1999-1-3
Johan Maljaarsa*, Mladen Lukićb, Frans Soetensc
a
TNO, Van Mourik Broekmanweg 6, 2628 XE Delft, Netherlands
b
CTiCM, 91193 Saint-Aubin, France
c
Eindhoven University of Technology, Den Dolech 2, 5612 AZ Eindhoven, Netherlands

Abstract

The fatigue behaviour of aluminium shows not only many similarities, but also some differences with the fatigue behaviour of
steel. There exist separate Eurocodes for the fatigue design of steel and of aluminium structures, EN1993-1-9 and EN1999-1-3,
respectively. The latter standard was developed after the draft version of the first one was available. However, a number of
aspects are considered in a different way in the two standards. Are these differences justified and desired? What can we learn
from the two standards for future improvement of the documents? This paper presents and discusses the agreements and
differences between the two Eurocodes. The paper evaluates the advantages and disadvantages of the approaches adopted in one
or another of the two standards. It may serve as a starting point for future harmonization of the two standards.

©
© 2013
2013TheTheAuthors.
Authors.Published
Publishedbyby
Elsevier Ltd.
Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection
Selectionand
andpeer-review
peer-reviewunder
underresponsibility of CETIM,
responsibility of CETIMDirection de l'Agence de Programme.

Keywords: Fatigue; steel; aluminium; welded joints; fracture mechanics;

1. Introduction

Steel as well as aluminium structures subjected to fluctuating loads may fail by fatigue cracking. The fatigue life
is characterised by an initiation period and a propagation period. Final fracture occurs when the remaining ligament
is no longer able to resist the loads. Fatigue cracks predominantly initiate at locations with high (local) stresses, such

* Corresponding author. Tel.: +31 88 8663464; fax: +31 88 8663016.

E-mail address: [email protected]

1877-7058 © 2013 The Authors. Published by Elsevier Ltd. Open access under CC BY-NC-ND license.
Selection and peer-review under responsibility of CETIM
doi:10.1016/j.proeng.2013.12.060
 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48 35

as geometrical discontinuities and flaws resulting from production. Hence, from a physical point of view, steel and
aluminium structures subjected to fatigue behave in a similar way.
The most important difference from an engineering point of view is that the fatigue strength of aluminium is
generally lower than that of steel. Especially the crack propagation life of aluminium components is lower, due to
the higher crack growth rate (approximately a factor 3 according to BS 7910:2005 [1], [2]) and the lower fracture
toughness at room temperature, the latter aspect resulting in a smaller fatigue crack at the moment of fracture.
Hence, the crack initiation period is relatively more important for aluminium than for steel components.
Both steel and aluminium components are predominantly joined by arc welding or bolting (or riveting). However,
other joining techniques such as adhesive bonding and friction stir welding (FSW) are more popular for aluminium
than for steel. The components themselves are predominantly fabricated by rolling in case of steel, and by extrusion,
rolling or casting in case of aluminium.
Because of the similarities between steel and aluminium structures subjected to fatigue, some engineering
standards such as the IIW recommendations [3], give rules and assessment methods for both steel and aluminium
structures. However, in the Eurocode system separate standards are provided for each material. Consequently there
exists a Eurocode for fatigue of steel structures – EN 1993-1-9 [4], hereafter called the ‘steel standard’ – and a
Eurocode for fatigue of aluminium structures – EN 1999-1-3 [5], hereafter called the ‘aluminium standard’. This
paper compares both standards. It focuses on the differences in the standards and it provides background for these
differences. The information in this paper may serve as a basis for future harmonization of the standards, which is
one of the goals of the CEN TC250 SC3 Evolution Group on EN 1993-1-9.
The first sections describe the similarities and differences between the two standards. At the end of the paper, the
differences are discussed and some ideas for harmonization are provided.

2. General comparison

In both standards the most extensively described assessment method is the method where the fatigue strength is
based on S-N curves – logarithmic relationships between the stress range and the number of cycles at which failure
occurs. In both standards, separate S-N curves are provided for each type of detail, but no distinction is made
between steel grades or aluminium alloys, between weld metals or between heat affected zones. An exception is the
aluminium alloy 7020, for which a higher fatigue strength is provided for parent metal. For both standards, the
fatigue strength is represented by the reference strength at 2 million cycles, Δσc. A constant amplitude fatigue limit
(CAFL) is assumed at 5 million cycles. Thus, both standards do not reflect the recent doubt on the existence of a
CAFL, [6]. In case of a variable amplitude load, a cut-off limit is defined at 100 million cycles, Fig. 1, and the
assessment is based on the linear damage accumulation rule of Palmgren-Miner.

¦ Nii
n
D (1)
i
where
ni is the spectral number of cycles with stress range Δσi, and
Ni is the number of cycles to failure for stress range Δσi.

A first, gross comparison of the standards shows that the aluminium standard provides more extensive
information. Considering the products and joining techniques described in the introduction, it is logical that the
aluminium standard provides the fatigue reference strength of adhesive bonded joints and of aluminium castings,
while the steel standard does not give this information. In this respect, it is worth noticing that the aluminium
standard does not give the fatigue reference strength of FSW joints. Tests on FSW joints have shown that the fatigue
strength is high, provided that the process parameters – such as clamping force and rotation speed – are properly
selected. Difficulties in the definition of the correct process parameters is one of the reasons why it is difficult to
provide a generally applicable fatigue reference strength for FSW joints in standards.
Other information that is provided in the aluminium standard but not – or less extensively – in the steel standard
is:
36 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48

x General information on failure mechanisms and potential sites for fatigue cracking;
x Stress derivation from Finite Element (FE) analyses;
x Qualitative information on weld improvement techniques;
x Fatigue strength reduction in case of exposure to sea water.

Fig. 1. S-N curve as defined in EN 1993-1-9 and EN 1999-1-3

3. Design concepts and safety levels

Both standards provide three design concepts, being the safe-life design, damage-tolerant design and design
assisted by testing. But the description and application of these methods contain important differences between the
standards.

3.1. Safe-life and damage-tolerant design

According to the steel standard, the safe-life method is to be used for details where local formation of cracks
could rapidly lead to failure, resulting in an unsafe character of relying on regular in-service inspection for fatigue
damage. Damage tolerant design in the steel standard may be applied when in the event of fatigue damage a load
redistribution between components can occur. It is to be used for inspected details that have a relatively low crack
rate, a long critical crack length and / or multiple load paths. In this case, a prescribed inspection and maintenance
regime for detecting and correcting fatigue damage should be implemented. According to the rules in EN 1993-1-10,
between 0 and 3 inspections are required during the life of the structure. The required reliability for a damage
tolerant design can be achieved by:
x selecting details, materials and stress levels so that in the event of the formation of cracks a low rate of crack
propagation and a long critical crack length would result;
x provision of multiple load path;
x provision of crack-arresting details;
x provision of readily inspectable details during regular inspections.

The difference in the assessment between safe-life and damage-tolerant elements in the steel standard is the
partial factor, refer to Section 3.2.
 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48 37

The steel standard does not provide methods to determine the inspection interval, nor the inspection method.
Definitions of ‘regular’ inspection, ‘low’ rate of crack propagation and ‘long’ critical length are not provided in the
standard. In many practical cases it may be difficult for the engineer to interpret ‘regular’, ‘low’ and ‘long’. In
addition the standard does not provide methods to determine the crack propagation rate and the critical crack length.
This may result in undesired differences in the design and maintenance of similar structures by different engineers
and owners.
The aluminium standard does not provide a distinction between elements suited for safe life and elements suited
for damage tolerant design. Both design methods can be used for all types of elements. In the aluminium standard,
safe life design is associated with damage assessment using the S-N curves. Damage tolerant design is associated
with the calculation of the crack growth using the fracture mechanics (FM) method. This method is used to
determine the inspection interval. The standard provides crack growth parameters for a number of alloys. The steel
standard does not provide the FM method; both safe-life and damage-tolerant elements should be assessed with the
S-N method according to this standard.
In a proposed amendment to the aluminium standard, [7], which is not yet a legislative document but may be
integrated into a next version of the standard, the damage tolerant method is extended to an assessment based on S-
N curves for those structures that either have sufficient redundancy to redistribute the load effects such that any
initiated crack propagation will stop, or have sufficiently large sections to carry the load effects after the first cracks
detectable by the naked eye have occurred. If these conditions are not fulfilled, the assessment with S-N curves is
complemented by a fracture mechanics calculation for inspection intervals. Only in this latter case, the design value
of the fatigue damage calculated with S-N curves is allowed to be larger than 1: The recommended limit value of the
damage is 2 in case of bolted, riveted or welded joints and 4 in case of parent metal. The amendment to the
aluminium standard provides more detailed information on the recommended time after completion of the structure
for the start of general inspections and fatigue inspections, as well as the required number of inspections.
Neither of the two standards provides information on the required detectable crack size or required probability of
detection. This may result in large and undesired differences in the reliability of similar structures assessed with the
damage tolerance method but maintained by different parties. By introducing the amendment, the two approaches
are more in line between the two standards, but still a significant difference remains in the prerequisites for usage of
the methods between the two standards.

3.2. Partial factors and damage limits

The partial factor for the fatigue load, γFf, is introduced in both standards. In case of steel, the recommended
value is specified in other Eurocodes such as EN 1993-2 for bridges. Its recommended value is in all cases γ Ff = 1,0,
but some National Annexes provide higher values. There is no aluminium equivalent for EN 1993-2 and therefore
the recommended values for γFf in case of aluminium are mentioned in EN 1999-1-3. In case the load spectrum is
taken from Eurocode parts EN 1991, the aluminium standard provides a recommended value γFf = 1,0. In addition,
the aluminium standard gives partial factors for cases where the fatigue load is based on measurements, Table 2.
Although the steel standard allows for using measurements for fatigue loads, the partial factor to be used in these
cases, or the derivation of the design value of the fatigue loads, is not provided.
According to both standards, the partial factors for the fatigue resistance, γ Mf, may be provided in the National
Annex. However, the recommended values differ significantly. The steel standard provides recommended values
according to Table 1, while the aluminium standard provides a value of γ Mf = 1,0 for all cases. This is an important
discrepancy between the standards, because the fatigue strength curves are based on a 95% probability of survival in
both standards and both standards make use of the EN 1991 parts for the fatigue load. Consequently, the safety
levels of the two standards are different.
For the aluminium standard, ref. [8] mentions that the reliability index over the entire design life in case of the
aluminium standard using γMf . γFf = 1,0 is in between 1,6 < β < 2,2, and that only in cases with partial factors γ Mf. γFf
>1,35 values of E | 3,5 may be reached, and this only at low distribution scatter. This means that the target
reliability in case of the aluminium standard is in many cases lower than the target reliability for fatigue according to
EN 1990. For resistance class 2, reference period 50 years, EN 1990 provides a target reliability of 1,5 to 3,8,
depending on the inspectability, reparability and damage tolerance. According to ref. [9] and [10] the reliability
38 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48

indices for steel with high failure consequence and steel with low failure consequence decrease during the life,
starting from approximately 10 and ending at 3,6 or 0,95 for high or low consequences, respectively, with all values
calculated for a design life of 100 years. If this is true, it means that the reliability index over the entire life is
importantly larger than the target reliability of EN 1990. The resulting reliability indices in ref. [9] and [10] are in
disagreement with ref. [11], where significantly higher partial values are derived for the target reliability indices
mentioned in EN 1990.
At this moment, neither of the standards complies with the distinction in consequence classes (CC) 1, 2 and 3
according to EN 1990. However, the evolution groups of both standards are in the process of drafting new
recommended values for γMf that are related to the CC. The evolution group of the aluminium standard is furthest
with this recommendation: in [7], partial factors γ Mf are provided that range between 1,0 and 1,1 for CC1, between
1,0 and 1,2 for CC2 and between 1,1 and 1,3 for CC3, where the highest values are for safe life design without any
inspection and the lowest values are for damage tolerant design. The exact values for γ Mf are at this moment under
discussion in the evolution group for the steel standard, but larger values than those in the current standard (Table 1)
are proposed, at least for safe life design. Hence, a discrepancy remains present between the safety levels of the
standards for safe life design if these values are adopted.

Table 1. Recommended values for partial factors in EN 1993-1-9 Table 2. Recommended values for partial factors in EN 1999-1-3
γFf Provided in other standards γFf 1,0 for fatigue load models in EN 1991
Not given in case of measuring
γFf (measure) 1,2 kn = 0 kn = 2
γMf Consequences of failure kf = 0 1,5 1,4
Assessment low high kf = 1 1,3 1,2
damage tol. 1,0 1,15 kf = 2 1,1 1,0
safe life 1,15 1,35
γMf 1,0 in all cases
1
Loads based on mean plus kf standard deviations
2
Number of cycles based on mean plus kn st. dev.

3.3. Measurements and tests

For cases where load data or fatigue strength data are not available, both standards provide the possibility to carry
out measurements or tests. Concerning tests for the fatigue strength, the steel standard simply refers to EN 1990 for
performing and evaluating the tests. However, EN 1990 does not specify how to derive a fatigue load spectrum, or
how to construct the design S-N curve from tests. The aluminium standard gives procedures to determine the fatigue
strength from test results, and a procedure to determine the design life in case of testing with a fatigue spectrum
representative for practice. One of the specifications is that the characteristic regression line is the lower value of the
mean minus two times the standard deviation and 0,8 times the mean. The latter value allows for wider variations in
production than is normally expected in a single set of fatigue specimens. However, aspects such as the required
stress levels to construct an S-N curve are also not provided in the aluminium standard.

4. Stresses and stress ranges

4.1. Stress orientation

In agreement with most other design standards for fatigue, the aluminium standard specifies that the principal
stress range should be used for weld toes, weld caps and parent metal. The principal stress range to be used is the
greatest difference between the principal stresses acting in planes of maximum 45° apart. If the direction of principal
stresses is more than or less than 45° to the weld axis it should be assumed to be normal to or parallel to the weld
axis, respectively. The steel standard specifies a different stress to be used. According to this standard, the direct
stress range and the shear stress range should be determined for most details. Separate S-N curves are provided for
direct and for shear stress ranges. The stress spectra should first be translated into equivalent stress ranges, ΔσE,2 and
ΔτE,2 respectively, giving the same damage at 2 million cycles as the entire spectra. Subsequently, the equivalent
stress ranges are combined in one damage calculation, Eq. (2). The relevant stresses in welds according to the steel
 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48 39

standard are according to Eq. (3), with stress components according to Fig. 2a. The aluminium standard specifies the
relevant stress for weld roots according to Fig. 2b. Fig. 2b. does not consider the relevant stress in case of a bending
moment or transverse shear force on the loaded attachment.

3 5
§ J Ff 'V E ,2 · § J Ff 'W E ,2 ·
¨ ¸ ¨ ¸ d 1,0 (2)
¨ 'V c J Mf ¸ ¨ 'W c J Mf ¸
© ¹ © ¹
'V wf V A f 2  W A f 2 and 'W wf 'W // f (3)

a b

Fig. 2. Stress ranges for welds; (a) according to EN 1993-1-9; (b) according to EN 1999-1-3

4.2. Type of stress

Both standards make use of nominal stress ranges, modified nominal stress ranges and hot-spot stress ranges The
definitions of these types of stress is more extensive and more exact in the aluminium standard. The interpretation is
left to the engineer in case of the steel standard. As an example, the modified nominal stress is the nominal stress
multiplied by an appropriate stress concentration factor to allow for a geometric discontinuity that has not been
taken into account in the classification of a standard detail. The aluminium standard specifies the types of
discontinuities to be considered and indicates for each detail which stress concentration effects are already covered
in the classification. This information is not provided in the steel standard. The steel standard provides hot-spot S-N
curves for the toes of regular types of welds. Contrary, the aluminium standard provides the hot-spot reference detail
method. In this method, the reference detail category, 'Vc,reff, should be selected form the table of standard details,
which is as similar as possible to the detail to be assessed. FE models should be developed for the detail to be
assessed as well as for the reference detail. The hot spot stresses VHS,assess and VHS,ref need to be determined from
these models for a unit load. The fatigue strength of the detail to be assessed, 'Vc,asses, is calculated according to Eq.
(4), refer to Fig. 3.

V HS,ref
'V c,assess 'V c,ref
V HS,assess
(4)
40 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48

Vref

Vasses VHS,ref

VHS,asses
Vc,ref = 36 MPa:

Fig. 3. Parameters to determine for the hot spot reference detail method

4.3. Effect of mean stress

The fatigue strength provided in the standards is based on a relatively high mean stress. A low, externally applied
mean stress may be beneficial for the fatigue strength, provided that residual stresses e.g. due to welding are low.
The steel standard indicates that the mean stress effect may be considered for non-welded or stress relieved details,
by calculating an effective stress range according to Eq. (5). For non-stress relieved welded joints, a mean stress
effect should not be considered in the steel standard. The aluminium standard gives a factor f( f R) with which the
fatigue strength may be multiplied. Contrarily to the steel standard, the aluminium standard also provides a factor
for simple welded joints. In that case, the stress ratio needs to be estimated including the effect of residual stresses,
according to Eq. (6). The factor according to the aluminium standard is provided with Eq. (7). The background to
this equation is provided in [12]. Fig. 4 compares the mean stress effect of the two standards, expressed as the factor
f(R) with which the fatigue strength may be multiplied. The figure shows that the mean stress effect allowed
according to the aluminium standard is larger than that of the steel standard.

1R
'V V max  0,6 ˜ V min
1  0,6R (5)
2V res  'V
R
2V res  'V (6)
­1 d 1,2  0,4R  d 1,6 plainmaterial
f (R) ®
¯1 d 0,9  0,4R  d 1,3 simpleweldedor bolted joints (7)
16
1.6 steel, non-welded
or stress relieved
15
1.5
alu, plain material
14
1.4
alu, simple welded
f(R)

1.3
1.3
or bolted joints
12
1.2
1.1
1.1
1
-1.5 -1 -0.5 0 0.5 1
R

Fig. 4. Effect of stress ratio (R = Vmin / Vmax) as defined in EN 1993-1-9 and EN 1999-1-3
 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48 41

5. S-N curves

5.1. Slope of the curves

Both standards give S-N curves per type of detail. The S-N curve is characterised by m1, m2 and Δσc (Fig. 1).
The aluminium standard specifies m1 and m2 per detail type, which reflects the slope of the tests. For each detail a
choice is made out of 4 slopes: m1 = m2 = 7 for plain members; m1 = m2 = 4 for bolted joints; m1 = 3,4 or 4,3
(whatever value fits the tests) and m2 = m1 + 2 for welded joints. By using one of these fixed slopes, it is less
difficult to provide the slope in case of a limited number of test results. In the steel standard, m1 = 3 and m2 = 5 for
all details loaded by direct stress except for tubular joints, and m1 = m2 = 5 for all details loaded in shear and for
tubular joints. This is an approximation; tests indicate that m1 is indeed approximately 3 for most welded details, but
that m1 is larger than 3 for details with a significant initiation life. As an example, Fig. 5 gives the S-N curve
according to the steel standard together with test results for preloaded bolted joints from [15]. For such details, a
differentiation in the slope would also be useful in case of steel, especially for non-welded details, as the curve with
m1 = 3 does not reflect the behaviour observed in tests. This may result in uneconomic designs, because the stress
ranges in practical structures are often smaller, and the numbers of cycles larger, than stress ranges and numbers
used in tests.

1000
stress range [N/mm2]

100

Joint with 8 bolts

Joint with 4 bolts
S-N curve EN 1993-1-9
10
1.E+3 1.E+4 1.E+5 1.E+6 1.E+7 1.E+8 1.E+9
no. of cycles [-]

Fig. 5. S-N curve for preloaded bolted steel joints in [4] with test results from [15]

5.2. Tables with detail types

Many types of details are used in both aluminium and steel structures, and so they are covered in both standards.
Examples are plain members, bolted joints, longitudinal welds, load carrying penetration joints, load carrying fillet
welds and non-load carrying welded attachments and stiffeners. Some tables are more extensive, with more types of
details, in the steel standard and some other tables are more extensive in the aluminium standard. There are detail
types covered in the steel standard but not in the aluminium standard. Examples are details in crane runway beams
and orthotropic decks (typical for steel) and lattice girder joints (should be assessed with the modified nominal stress
or hot-spot stress method in case of aluminium). Finally the aluminium standard gives separate tables for welded
attachments on built-up beams (plate girders) and for welded attachments on other members.

5.3. Execution levels

General execution requirements and execution levels for steel and for aluminium structures are provided in EN
1090-2 and in EN 1090-3, respectively, with different requirements for so-called execution classes. There is no
42 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48

principal difference between these execution standards. In the fatigue standard for aluminium, different S-N curves
are provided for different execution classes, Fig. 6. The steel standard requires a general execution class and gives
no differentiation per detail type. In addition, both the steel and the aluminium standard provide some additional
requirements for a number of details. As an example, a higher fatigue strength is obtained when the weld is ground
flushed, Fig. 6.

Fig. 6. Example of a table from EN 1999-1-3 with differentiation in execution level

5.4. Fatigue reference strength

A typical difference between the standards is that the aluminium standard provides thickness and length effects
only for a small number of details. As an example, thickness effects are not considered for butt welds with caps and
length effects are not considered for some types of welded attachments. With fracture mechanics it can be proved
that thickness and length effects are present in such types of joints. These effects are considered in the steel
standard.
In case of fillet or partial penetration welded joints, the aluminium standard specifies that toe cracking should be
checked when the weld throat a > 0,6 times the plate thickness, t, and root cracking should be checked when a d 0,6
t. Considering the respective fatigue strengths ('Vc = 28 and 25 N/mm2, respectively, this is a logic division in case
of loading by a normal force. However in case of loading by a bending moment and / or transverse forces this
division is not correct. The steel standard gives a general description; toe and root cracking should be checked in
case of fillet and partial penetration welds. With this description all cases are covered.
Fig. 7 gives the ratio between the reference fatigue strength, ΔVc, according to the aluminium standard and Δσc
of the equivalent detail in the steel standard. Not all details are covered, because for some details an equivalent
detail in the other standard is missing. Furthermore, the equivalent details are selected in such a way that they match
as much as possible, but deviations may occur, for example due to differences in execution requirements and
thickness and length effects. The figure shows that the ratio for most parent metal details is approximately equal to
0,5 and for most welded joints the ratio is between 0,35 and 0,5. The test data of details with a ratio that significantly
deviates from these ratios could be re-evaluated in order to determine whether the fatigue strengths in the standards
are correct for these details. A first, gross estimate is provided here for these details, based on theoretical principles,
comparisons with other standards and comparisons with FM calculations:
a) Non-preloaded bolted joints (steel detail 8.1-13 and aluminium detail 15.2).
According to the steel standard, the ratio of Δσc between preloaded and non-preloaded bolted joints is 2,2, and
the stress should be based on gross and net cross-section, respectively. This is in agreement with other
 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48 43

standards; for example this ratio is 2,3 in the US standard for fatigue of steel structures, Erreur ! Source du
renvoi introuvable.. In the aluminium standard, the ratio is significantly different: it is equal to 1. In addition, a
discrepancy exists in the aluminium standard in reference fatigue strength and slope of the S-N curve between
plain elements with a round hole (m1 = m2 = 7, Δσc = 42 N/mm2) and non-preloaded bolted joints (m1 = m2 = 4,
Δσc = 56 N/mm2). From a physical point of view, there is no difference in fatigue strength between these two
details. The category for non-preloaded bolted joints in the aluminium standard needs re-evaluation.
b) Welded attachments (steel details 8.4-6, 7 and 8 and aluminium detail 3.8).
The relatively low fatigue strength of the aluminium connection may be due to the fact that length effects are
not considered in the aluminium standard: in Fig. 7 the strength is compared with Δσc = 80 N/mm2 for an
attachment in steel with length < 50 mm.
c) Toe failure of fillet welds (steel detail 8.5-1 and aluminium detail 9.1).
In the steel standard, there is no distinction in Δσc for toe cracking of full penetration and of fillet welded joints.
According to the aluminium standard, the ratio between Δσc for toe cracking of full penetration and fillet
welded joints is equal to 1,3. Considering the local stress distribution it is logical that the fatigue strength of toe
cracking fillet welded joints is lower than that of full penetration joints. With FM an estimate of the ratio of the
fatigue strength of full penetration and fillet welded joints is obtained, which is approximately equal to 1,5. The
fatigue strength of toe cracking fillet welded joints in the steel standard is expected to be too high for fillet
welded joints. For joints loaded with a normal force, this has no implication, because these joints normally fail
by root cracking. This may be different in case of loading by a bending moment, because bending may result in
a larger ratio between stress at the weld toe and stress at the weld root as compared to normal force.
d) Root failure of fillet welds (steel detail 8.5-3 and aluminium detail 9.2).
In the steel standard the ratio of Δσc between root cracking of fillet welded joints and of partial penetration butt
welds is equal to 1,0. In the aluminium standard, this ratio is equal to 1,4. With FM – using the solutions and
crack growth parameters from BS 7910 – one can provide an estimate of Δσc. Depending on the selection of
variables, the resulting values of Δσc are approximately 32 N/mm2 and 11 N/mm2, for steel and aluminium,
respectively. The values are more or less in agreement with EN 1993-1-9 (Δσc = 36 or 40 N/mm2) but in
disagreement with EN 1999-1-3 (Δσc = 25 N/mm2). The fatigue recommendations of IIW, Erreur ! Source du
renvoi introuvable., provides values of Δσc = 36 N/mm2 for steel and Δσc = 12 N/mm2 for aluminium. It is
recommended that the provided fatigue strength in the aluminium standard be checked.

The aluminium standard provides the fatigue strength of details exposed to sea water, to be taken two detail
categories lower than the fatigue strength exposed to air. The steel standard does not provide the fatigue strength of
details when exposed to sea water. This is an omission for the design of structures such as lock doors.
44 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48

1.2 parent longitudinal transverse welded load carrying

'Vc,alu / 'Vc,steel
1 metal welds butt welds attachments welded joints

0.8

0.6

0.4

0.2

0
alu

7.3.2
7.3.1
7.2.2

7.4.1
7.1.1
7.1.2
7.2.1
15.1

3.18
3.18

13.5
11.4
15.2

5.1
5.3

7.5

7.6
1.2

3.7

3.8
7.6

9.5
1.4

5.3
5.4
5.6
5.7

3.6
3.5

9.1
9.2
9.4
detail

8.4-4
8.1-1-3
8.1-5

8.2-4
8.2-5
8.2-8
8.2-9

8.3-17

8.4-3

8.4-6-8
8.4-5

8.5-1

8.5-5
8.5-1
8.5-3
8.5-4
8.2-10
8.2-11

8.3-13

8.4-1
8.4-1
8.1-8

8.3-11

8.5-6
8.1-13

8.3-14+15
8.3-14+15
8.3-2+3
8.3-1+4

8.3-5+7
8.3-6

8.3-13
steel
detail
a)

b)) c)) d)

Fig. 7. Ratio between Δσσc for an aluminium detail and Δσσc for the equivalent steel detail

6. Fracture toughness

The fatigue life ends when fracture occurs and thus the fatigue life depends on the fracture toughness. The
fracture toughness varies between steel grades, aluminium alloys and heat affected zones but also between batches
of the same steel grade or aluminium alloy. It is well known that the fracture toughness of steel may be very low in
case of low temperatures. This aspect is covered by the thickness requirements in EN 1993-1-10, which cause that a
steel structural component has a certain minimum toughness, preventing it from premature failure and a short
fatigue life. Toughness requirements and toughness tests are not prescribed for aluminium by EN 1090-1 nor by the
EN 1999 parts, because the fracture toughness generally increases at decreasing temperatures. However, the fracture
toughness of aluminium at room temperature is relatively low. Values ranging between KIC = 22 MPa—m and KIC =
50 MPa—m for alloys of the same group but with other tempers are reported in [18]. Similar values are provided in
[19].
In order to explore the influence of the variation in the fracture toughness of aluminium alloys, a fracture
mechanics calculation is carried out on a plate with a welded attachment, with plate thickness T = 25 mm, node
length L = 25 mm, stress range 'V = 18 MPa, semi-circular crack with initial depth ai = 0,15 mm – representative
for an as welded structure – and crack growth parameters for aluminium according to [1]. The calculations are
carried out for two selected values of the fracture toughness for aluminium alloys, and for two maximum stress
values, equal to the 0,2 % proof stress and half that value. The results provided in Table 3 indicate that the critical
crack depth, acr, and the number of cycles up to that critical crack depth, Nf, depend very significantly on the
fracture toughness for these low values.
 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48 45

Note that fracture mechanics remains a model for crack propagation only, and the real number of cycles and
crack depth may deviate from the values in Table 3. Nonetheless, the large influence which is found here forms a
potential risk, because the tests upon which are developped the S-N curves in the aluminium standard are generally
based on a limited number of alloys, while they may be applied to other types of alloys in practice. Moreover the
critical crack depth may be so low in certain circumstances, that crack detection by in-service inspection is not
feasible.

Table 3. Results of fracture mechanics calculations of an aluminium plate with welded attachment (T = 25 mm, L = 25 mm,
ai = 0,15 mm, ci = 0,15 mm, 'V = 18 MPa)
material K1C = Toughness [MPa—m] 22 50
loading Vmax / f02 = ratio max. stress / proof stress [-] 1 0,5 1 0,5
results acr = critical crack depth [mm] 0,16 0,8 0,8 4,1
acr / T 0,0064 0,0320 0,0320 0,1640
Ncr = number of cycles up to acr [x106] 0,06 1,7 1,8 3,1

7. Recapitulation and discussion

Descriptions of the similarities and differences in the formats of the two standards are provided in the previous
sections. This section discusses the differences, gives the opinion of the authors to which standard is better suited for
what, and gives a number of ideas for future harmonization.
A number of aspects mentioned in the previous sections are relatively easy to harmonise and modify in one or
both standards. This applies to:
x The stress definition for welds (Section 4.1). The stress in aluminium standard does not cover all possible load
cases in the definition of the relevant stresses in the weld. A format such as the one used for the steel standard
may be implemented to overcome this problem.
x The effect of the mean stress (Section 4.3). Although differences between the two materials may be present in
the effect of the mean stress on the fatigue life, the format of the equations can be equal. Eq. (6) provides a
conservative approximation of the stress ratio, R, because the residual stress is multiplied by 2. Instead, it would
be more suited to account for uncertainties in the residual stress in the relationship f(R) of Eq. (7).
x The slope of the S-N curves (Section 5.1). A differentiation in slope for welded and non-welded details in the
steel standard might be desirable in order to prevent uneconomic designs. On the contrary, this might bring
slightly more complexity to the standard and thus to the structural analysis.
x Differentiation of the fatigue strength with respect to the execution class (Section 5.3). Differentiation provides
a choice between easy execution level but low fatigue strength on the one hand, and high execution level and
high fatigue strength on the other hand, and is therefore useful. A differentiation similar to that in the
aluminium standard may also be adopted in the steel standard. Complicating factor in the generation of S-N
curves for different execution levels is that test results need to be available, and that the execution requirements
in EN 1090 do not fit one to one with the fatigue strength. Hobbacher and Kassner [17] have provided a list of
execution requirements for fatigue welded details. This list can be used to provide the execution requirements
for the details in the steel standard.
x Re-evaluation of a number of detail categories in the steel and in the aluminium standard, and the lack of
thickness and length effects for some details in the aluminium standard (Section 5.4).
x Implementation of S-N curves for the details when exposed to (sea) water (Section 5.4). There is sufficient
information from literature to derive these S-N curves for steel and to implement it in the standard – as already
carried out for the aluminium standard.
x The illustrations of the details (Section 5.4). The illustrations in the aluminium standard are in general more
uniform, clearer and more unambiguous as those in the steel standard.
46 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48

For a number of aspects, harmonization requires more research and discussion between the evolution groups of
the two standards. The authors of this paper provide some ideas for proceeding with these issues in the following
sections.

7.1. Design concepts

Section 3.1 of this paper provided the differences between the safe-life design and damage-tolerant design
methods between the standards. A number of aspects and definitions are ambiguous, especially in the steel standard
but also in the proposed amendment to the aluminium standard. Examples are the lack of definitions of ‘regular’
inspection, ‘low’ rate of crack propagation and ‘long’ critical length. The steel standard does not provide methods to
determine the crack propagation rate and both standards do not provide the critical crack length. Finally, the
standards do not provide guidance on the detectable crack size or the required probability of inspection, which may
result in undesired differences in the design and maintenance of similar structures by different engineers and
owners. There are various possibilities to solve this matter:
x Provide additional guidance via rules, notes and examples on inspection intervals, crack propagation rates and
critical crack lengths for various types of structural details.
x Provide the FM method, especially focusing on relatively large cracks, so that the engineer is able to calculate
crack propagation rates and critical crack lengths, and so that he can determine the effect of the detectable crack
length on the inspection interval.

The problem with the FM method is that it requires high-level skills – and a substantial amount of work – to
incorporate aspects such as the beneficial effect of multiple load paths and constraining effects of adjacent material
in the calculation of the crack growth rate and the critical crack length. Ignoring these effects may result in
conservative estimates and therefore in impractically short inspection intervals. This pleads for the first possibility.
A noticeable difference between the FM method and the S-N curves in the aluminium standard is that the crack
growth rate is specified per alloy, while the S-N curves are independent of the alloy. Considering the relatively small
differences in crack growth rates in the standard, it seems possible to provide only one rate, independent of the alloy.
This also overcomes the impracticability that the crack growth rate of a number of alloys is currently not provided in
EN 1999-1-3, while these alloys are allowed in structural applications according to EN 1999-1-1. Moreover, the
difference in crack growth curves between parent material and the heat affected zones are expected to be larger than
the differences between alloys. Another important impracticability with the FM method in the aluminium standard is
that the material toughness is not provided, nor are methods prescribed how to determine the toughness and / or the
critical crack length. This is an omission because Section 6 indicated that even for the entire life, the fracture
toughness of aluminium already has a dramatic impact on the fatigue life. This is even more important in case of the
residual life – the life between an inspection and final fracture of an element. Regarding the relatively small critical
crack sizes calculated in Section 6, it is questionable whether it is feasible to base inspection intervals on fracture
mechanics calculations as implemented in the aluminium standard, especially for elements without multiple load
paths. To summarize, the FM method in the aluminium standard is not yet completely elaborated and requires
updating in a new version of the standard.
Regarding the third design concept, design assisted by testing, the aluminium standard gives more information on
how to carry out tests and how to construct a design S-N curve from tests. This information could also be added to
the steel standard. In view of the authors, it would be valuable to add information on the different load levels to be
applied in tests in both standards, so that the resulting S-N curve is not based on a cluster of test results with almost
identical load level.

7.2. Partial factors

Related to the design concepts is the issue of correct values for the partial factors, Section 3.2. In view of the
authors, the amendments for both standards in which the partial factors are based on the consequence class are a
good way forward, because it prevents interpretation differences of ‘high’ and ‘low’ consequences of failure of a
 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48 47

component. It is obvious that it is desirable to provide recommended partial factors in the two standards that are in
line with each-other, which is currently not the case. Starting point is a uniform selection for the reliability index in
case of fatigue. In this respect, one could either take the target reliability index in EN 1990 as a starting point, or it
could be considered that there is no – or only a limited – economic value of the structure at the end of the service
life. In the latter case, only human safety considerations play a role in the target reliability index. Ref. [20] provides
corresponding target reliability indices for that case. The required partial factor γMf cannot be seen independent of
the definition of the fatigue load, including the partial factor on the load side and the coefficient of variation of the
load. Setting correct values for the partial factors needs an additional study involving the evolution groups of EN
1993-1-9, EN 1999-1-3 and various parts of EN 1991.

An omission in the steel standard is the partial factor, and the derivation of the characteristic fatigue load
spectrum, in case the fatigue load spectrum is based on measurements. The aluminium standard gives guidance for
this. If the partial factors are agreed by all parties, this could easily be implemented in the steel standard.

7.3. Stress derivation

Section 4 showed that there is a difference in the stress to be used in the verification. The aluminium standard
uses principal stresses, while the steel standard uses direct and shear stresses, and combines these stresses in the
verification. From a theoretical point of view, there is no reason why the stress type is different for steel and
aluminium components. In most cases with combinations of direct and shear stresses, the procedure based on the
principal stresses results in a (slightly) larger value for the calculated damage. More information and test data is
required in order to determine the concept that best reflects the real behavior in case of complex loading, both in and
out of phase. But in most practical cases either direct or shear stress dominates so that there is no difference between
the two methods.

For the hot-spot method, the steel standard provides a table with hot-spot S-N curves, while the aluminium
standard uses the fatigue reference method. The advantage of this method is that the method is less depending on the
mesh type, mesh size and exact modeling of the joint. The disadvantage is that the selection of an appropriate
reference detail may be problematic in some cases. Both standards do not specify how the hot-spot stress should be
determined, i.e. the extrapolation method is missing. This definition can and should be added to new versions of the
standards. Apart from this omission, the aluminium standard gives extensive information on the use of FE models in
fatigue design, and how to extract stresses from FE models for the application in the fatigue check. The steel
standard does not give this information. Some aspects mentioned in the aluminium standard are useful, such as the
warning that thin shell elements cannot model the real stiffness and stress distribution inside, and in the vicinity of,
the weld zone of intersecting shells. But most of the information is so general that it is of ample use for the engineer
in practice and that it is unclear why the information is provided in a standard for fatigue.

7.4. Fracture toughness

From Section 7 it is evident that the current absence of a requirement on the fracture toughness in Eurocode 9 is
an omission and a risk for fatigue loaded aluminium structures. In a future version of the standard, requirements on
the minimum fracture toughness should be provided in order to meet the fatigue life calculated with EN 1999-1-3.

8. Conclusions

This paper compares and evaluates the Eurocodes for fatigue of steel and aluminium structures, EN 1993-1-9 and
EN 1999-1-3, respectively. The same physical principles apply to the fatigue behavior of steel and aluminium
components, but the standards contain a number of significant differences. This paper shows that, for most
differences between the standards, there is no reason why they need to be maintained. For these aspects, one concept
can be applied in future versions of the standards without substantial additional research. The most important
aspects to which this applies are:
48 Johan Maljaars et al. / Procedia Engineering 66 (2013) 34 – 48

x Stress definition for the assessment of welds.

x Effect of the mean stress.
x (Possibly) slope of the S-N curves.
x Differentiation of the fatigue strength with respect to the execution class.
x Re-evaluation of the reference fatigue strength of a number of details that appear not in line between the two
standards.
x Implementation of S-N curves for the details exposed to (sea) water.
x Illustrations of the details.

Issues that need to be worked on for future versions of both standards are:
x Clear definitions and distinctions between safe life and damage tolerant design, and guidance on inspection
methods or detectable crack sizes and required probabilities of detection of inspection techniques need to be
provided, so that interpretation differences by designers are prevented.
x Recommended partial factors for the fatigue strength that are in line in both standards need to be derived, with
the target reliability values in EN 1990 and with the fatigue loads in various parts of EN 1991.
x The definition of nominal and hot-spot stresses. This applies to the combination of direct and shear stress or
principal stress, and the definition of the hot-spot stress.

A number of aspects may remain different between the standards, because of different physical behaviour or
different structural applications. The most important issue to which this applies is the difference in fracture
toughness. The fracture toughness of steel decreases with decreasing temperature, and the rules in EN 1993-1-10
prevent applications with brittle material. Instead, the fracture toughness of aluminium alloys is relatively low at
room temperature, and requirements on minimum toughness values are currently missing. This strongly influences
the fatigue life and the possibility of assessments using inspection results. It is important that requirements on the
minimum toughness be implemented in a future version of Eurocode 9 or EN 1090-3.
Harmonisation of the steel and aluminium Eurocodes is important, because this allows easy changing between
steel and aluminium for engineers that are used to work with either of the materials. Thus, it removes barriers and
allows for an optimal choice of materials. The evolution groups of the two standards will be working towards
harmonisation of the two standards.

References

[1] BS 7910:2005 Guide to methods for assessing the acceptability of flaws in metallic structures, BSI, 2005.
[2] Maddox, S.J. Int. J. of Fatigue 25, 2003, 1359-1378.
[3] Hobbacher, A., IIW doc. XIII-1965-03 / XV-1127-03 Recommendations for fatigue design of welded joints and components, 2005.
[4] EN 1993-1-9:2006 Eurocode 3: Design of steel structures – Part 1-9: Fatigue, CEN, Brussels, 2006.
[5] EN 1999-1-3:2007 Eurocode 9: Design of aluminium structures – Part 1-3: Structures susceptible to fatigue, CEN, Brussels, 2007.
[6] Pyttel, B., Schwerdt, D., Berger, C. Int. J. of Fatigue 33 (1), 2011, 49-58
[7] Draft amendment to EN 1999-1-3, 2011-06-17, SC 9 doc. N347
[8] Kosteas, D., Proc. AlFat EC9 Brussels, 2008.
[9] Nussbaumer, A. Borges, L., Davaine, L. Fatigue design of steel and composite structures, ECCS, Brussels, 2011.
[10] Sedlacek, G., Leitfaden zum DIN-Fachbericht 103 „Stahlbrücken“, 2003, 107-129
[11] Sørensen, J.D. et al. Baggrundsundersøgelser ifm. udarbejdelse af Nationale Annekser til EN1990 og EN1991, 2003
[12] Kosteas, D., American Society for Testing and Materials, 1988, 122-130.
[13] Fatigue tests on steel bolts, Offshore Technology Report OTO 97067, Health & Safety Exc, London, 1998.
[14] ANSI/AISC 360-10, Steel Construction Manual, Specification for Structural Steel Building, Appendix 3: design for fatigue, 14th ed., AISC,
2010.
[15] Hobbacher, A. Kassner, On relation between Fatigue Properties of Welded Joints and Quality Groups in ISO 5817, IIW doc. XIII-2323-10.
[16] Kaufman, J.G., Moore, R.L., Schilling, P.E., Engineering Fracture Mechanics, 1971 (2), 1971, 197-210.
[17] Koçak et al. Fitnet – European Fitness for Service Network
[18] Steenbergen, R.D.J.M., Vrouwenvelder, A.C.W.M, Heron 55 (2), 2010, 123-140.

V i e w p u b l i c a t i o n s t a t s