Fatigue S-N curves of bolts and bolted connections for

application in civil engineering structures

Citation for published version (APA):
Maljaars, J., & Euler, M. (2021). Fatigue S-N curves of bolts and bolted connections for application in civil
engineering structures. International Journal of Fatigue, 151, [106355].
https://doi.org/10.1016/j.ijfatigue.2021.106355

Document license:
CC BY

DOI:
10.1016/j.ijfatigue.2021.106355

Document status and date:

Published: 01/10/2021

Document Version:
Publisher’s PDF, also known as Version of Record (includes final page, issue and volume numbers)

Please check the document version of this publication:

• A submitted manuscript is the version of the article upon submission and before peer-review. There can be
important differences between the submitted version and the official published version of record. People
interested in the research are advised to contact the author for the final version of the publication, or visit the
DOI to the publisher's website.
• The final author version and the galley proof are versions of the publication after peer review.
• The final published version features the final layout of the paper including the volume, issue and page
numbers.
Link to publication

General rights
Copyright and moral rights for the publications made accessible in the public portal are retained by the authors and/or other copyright owners
and it is a condition of accessing publications that users recognise and abide by the legal requirements associated with these rights.

• Users may download and print one copy of any publication from the public portal for the purpose of private study or research.
• You may not further distribute the material or use it for any profit-making activity or commercial gain
• You may freely distribute the URL identifying the publication in the public portal.

If the publication is distributed under the terms of Article 25fa of the Dutch Copyright Act, indicated by the “Taverne” license above, please
follow below link for the End User Agreement:
www.tue.nl/taverne

Take down policy

If you believe that this document breaches copyright please contact us at:
[email protected]
providing details and we will investigate your claim.

Download date: 25. Jul. 2023

 International Journal of Fatigue 151 (2021) 106355

Contents lists available at ScienceDirect

International Journal of Fatigue

journal homepage: www.elsevier.com/locate/ijfatigue

Fatigue S-N curves of bolts and bolted connections for application in civil
engineering structures
Johan Maljaars a, b, *, Mathias Euler c
a
Eindhoven University of Technology, Groene Loper 3, Eindhoven, the Netherlands
b
TNO, Stieltjesweg 1, Delft, the Netherlands
c
Brandenburg University of Technology, Institute for Construction Engineering, Cottbus - Senftenberg, Germany

A R T I C L E I N F O A B S T R A C T

Keywords: Bolts and bolted connections are frequently used in civil engineering steel structures. This paper presents a meta
Fatigue tests study where a few thousand fatigue tests on these elements are evaluated. The evaluation reveals that current
Bolts specifications in design standards need updating to account for the relevant stress parameter and production
Bolted connections
methods. This substantially reduces the scatter of the fatigue resistance. The shape and position of the fatigue
Effect of preload
Effect of hole forming
resistance (S-N) curves also require updating. The results of this study have been implemented in the new
EN 1993-1-9 revision of European standard EN 1993-1-9. This paper provides the background for the modifications.

modifications of the FAT classes, of the slope parameter, and for some
details of the definition of the relevant stress for fatigue based on new
1. Introduction
insights obtained by this evaluation. This paper presents the background
of the newly derived FAT classes.
Bolts and bolted connections are frequently applied in civil engi­
neering structures such as bridges, crane supporting structures, masts,
2. Evaluation procedure
towers and chimneys. If carefully produced, some of these connections
may have a better performance in fatigue as compared to welded con­
2.1. Selection of materials and production methods
nections. Design standards and recommendations provide FAT classes
for the design of such products and connections. The FAT class refers to
Materials and production methods are selected that are typically
the 95% exceedance fraction of the fatigue resistance in [MPa] at two
used in civil engineering structures. Surface treatments such as polishing
million cycles that can be used in the design.
are usually not applied in this field. Mill scales were thus not removed in
Table 1 presents the FAT classes of bolts and of the most frequently
the specimens selected. An exception are DCC with preloaded high
applied bolted connections in three (inter) nationally used design stan­
strength friction grip bolts, where a surface treatment such as blasting is
dards, namely the American Bridge Design Specifications AASHTO:2012
applied at the contact surfaces, both in practice and in the selected
[1], the European Standard EN 1993-1-9:2005 [2] and the British
specimens. Only full size specimens were selected, with plate thickness
Standard BS 7608:2014 [3]. Double Covered Connections are abbrevi­
between 9 mm ⩽ t ⩽ 30 mm and bolt diameters between 12 mm ⩽D⩽ 72
ated to DCC in this table. In all cases, a linear relationship is assumed
mm.
between the logarithm of the applied stress range and the logarithm of
The fatigue resistance may be correlated to the grade of steel. Some
the number of cycles to failure in the finite life region. The slope
of the used sources report an increased fatigue resistance for steels of
parameter is m = 3 for all [1,2] or most [3] details, i.e. an endurance
higher grade. Others, however, did not observe such an influence. Fig. 1
equivalent to the reciprocal of the stress range to the power of three.
gives an illustrative example. The data from [4] for a 15 mm thick plate
Fatigue test data have been collected and evaluated for updating the
with an oxy-fuel cut hole shows no dependency on the steel grade
FAT classes in the revision process of the European standard EN 1993-1-
(subfigure a), whereas the opposite is found for plasma cut holes (sub­
9. Thousands of test data have been collected, of which almost two
figure b). This is not necessarily related to the cutting process; the same
thousand are considered relevant for civil engineering structures and are
authors did not find a consistent relation between fatigue resistance and
used to derive FAT classes. This meta study has resulted in substantial

* Corresponding author at: Eindhoven University of Technology, Groene Loper 3, Eindhoven, the Netherlands.
E-mail address: [email protected] (J. Maljaars).

https://doi.org/10.1016/j.ijfatigue.2021.106355
Received 12 April 2021; Received in revised form 12 May 2021; Accepted 27 May 2021
Available online 2 June 2021
0142-1123/© 2021 The Author(s). Published by Elsevier Ltd. This is an open access article under the CC BY license (http://creativecommons.org/licenses/by/4.0/).
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

Nomenclature σu Tensile strength

σy Yield stress
Abbreviations ξ Stress ratio correction factor for Morrow’s equation
6PRFLM Six parameter random fatigue limit model ζ Stress ratio correction factor for Soderberg’s equation
DCC Double covered connection a Constant in Basquin’s relation
DOF Degree of freedom a′ Constant in the 6PRFLM
FEM Finite element method b High cycle exponent in Morrow’s equation
FM Fracture Mechanics c1 , c2 , c3 Coefficients in the SCF for DCC with non-preloaded bolts
HDG Hot-dip galvanized D Bolt diameter
HT/R Bolts heat treated and then rolled d0 Hole diameter
R/HT Bolts rolled then heat treated f Confidence factor
RO Runout k Number of bolt rows per side of the connection
SCF Stress concentration factor Kf Fatigue notch factor
kn Prediction bound factor
Operators
Kt Stress concentration factor
∧
Estimator byp
Kt Stress concentration factor for bypass loading
Symbols Kpin
t Stress concentration factor for pin loading
Δσ C,M30 , Δσ C,M36 fatigue reference resistance at N = 2⋅106 of a 30 or m Slope parameter in Basquin’s relation
36 mm diameter bolt m′ Slope parameter in the 6PRFLM
Δσ C fatigue reference resistance at R = 0.5 and N = 2⋅106 N Number of cycles to failure
Δσ lim Constant amplitude fatigue limit n Number of failed tests
Δσ R,mod Modified stress range at stress ratio R nRO Number of run-out tests
Δσ R,net Net section stress range at stress ratio R p Slope transition parameter in the 6PRFLM
Δσ R Stress range at stress ratio R q Number of future samples
ΔτC Fatigue reference shear resistance at N = 2⋅106 R Stress ratio
ΔτR Range of average shear stress at stress ratio R s Estimate of standard deviation of log(a)
η Stress ratio correction factor for Walker’s equation t Plate thickness
σ′f High cycle fatigue resistance in Morrow’s equation t0.05 5% coefficient of Student’s T distribution
σm Mean stress w Ratio between plate width and number of bolts per row
σR,max Maximum stress of a cycle with stress ratio R

steel grade for 15 mm thick plates with plasma cut edges in [5]. It is also [10,11]. One reason for these different observations is the thickness of
not limited to thermal processes; different results regarding the influ­ the zinc layer; an influence is expected for a zinc layer exceeding a
ence of the steel grade are also reported for details such as DCC with certain thickness [12], as the microcracks in the zinc layer are then deep
preloaded bolts. Any influence of the steel grade may thus be depending enough to exceed the intrinsic threshold of the stress intensity factor
on the production process and the type of detail, but it is not the entire range of the base metal. HDG steel is generally not considered in the
explanation. current work, except for specific details and series where HDG test re­
An important reason for these different observations may be in the sults are used to confirm a general trend in case the database of non
surface quality of the specimens. Gurney [6] observed that the fatigue galvanized samples was small. Details will be given in the subsequent
resistance increases with increasing tensile strength for plain machined sections. Bolts are again the exceptions, for which a separate evaluation
specimen. In case of (sharply) notched specimen, the fatigue resistance was done on galvanized specimens.
did not increase with the tensile strength. The influence of the steel
grade, if any, is also related to the mean stress of the cycle [7]. 2.2. Influence of the stress ratio
The fatigue resistance may thus benefit from a high steel grade in
plain material and mechanically fastened connections in practical ap­ It is well known that the fatigue resistance of the considered
plications, provided that the specimen surface is relatively smooth. constructional details depends on the mean stress. Three types of
Given the treatment of materials during construction and use, an generally accepted and often applied mean stress corrections have been
initially smooth product cannot always be guaranteed to remain smooth considered for comparing and pooling the fatigue test data that were
during the entire life without special provisions. Most civil engineering carried out at different mean stress values, σ m . The equations are
structures are currently constructed from steel grades with a nominal generally applicable to tension-tension and tension-compression cycles
yield stress below 500 MPa. The S-N curves derived per detail in the (i.e. excluding compression-compression cycles).
following sections are therefore evaluated for steel grades with a nom­ The first mean stress correction is Morrow’s correction on the Coffin-
inal yield stress between 235 MPa ⩽ σy ⩽ 460 MPa. This includes the Manson relationship. The high-cycle fatigue part of the equation,
European grades S235, S355 and S460, the American grades A36, A441 considering stress ranges ΔσR that remain in the linear elastic stage,
and A572, and equivalent steel grades. Data on higher steel grades are reads:
occasionally used to determine the influence of certain variations, such ( ′ )
as the mean stress. Bolts are the exceptions, for which grades up to 10.9 ΔσR = 2(2N)b σ f − σm (1)
(nominal yield stress of 900 MPa) have been considered.
Different observations have been reported on the influence of the where N is the number of cycles to failure and σ f and b are resistance
′

zinc layer in Hot-Dip Galvanized (HDG) steel. Some authors show a clear
parameters. If correcting the exponent b for the surface roughness of the
reduction of the fatigue resistance of the detail types mentioned in the
mill scale according to [7] or [13], it appears that the fatigue resistance
introduction [8,9] whereas others do not show a significant influence
is almost independent of the material tensile strength around the fatigue

2
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

limit. Eq. (1) then results into the following stress ratio dependency: only depending on the stress range, but also on the maximum stress,
σ R,max . The Walker equation reads:
1− R
Δσ R = Δσ 0 (2)
1 − ξR (6)
η
) η η
Δσ(1−
R σ R,max = Δσ −(1−1 ) σ η− 1,max
where R is the stress ratio according to Eq. (3), Δσ R is the stress range at This equation is a generalised form of the Smith-Watson-Topper
stress ratio R (hence Δσ0 is the stress range at R = 0) and ξ = parameter [18] written in the shape of Langlas and Vogel [19], where
1 − Δσ0 /σf . The format of Eq. (2) is applied in many old German stan­
′
η = 0.5 in the high cycle range. Eq. (6) can be written as:
dards, such as [14] for riveted railway bridges, where ξ ranges between ΔσR = Δσ 0 (1 − R)η (7)
0.4 and 0.6.
Generally, a detail dependent Stress Concentration Factor (SCF)
σm − 0.5Δσ R applies, which implies that ξ, ζ, and η depend on the type of detail. All
R= (3)
σm + 0.5Δσ R collected fatigue test data have a stress ratio ranging between
Soderberg [15] relates the fatigue resistance to the ratio between the − 1⩽R⩽0.5. With the flexibility introduced by ξ, ζ, and η and for the
applied mean stress and the material yield stress, σy : considered ranges of steel grade and stress ratio, Eq. (2), (5) and (7) can
reasonably approximate other stress ratio dependencies, such as those of
Δσ R σ m Goodman [16], Gerber [20] and Dietmann [21]. The dependency of
+ =1 (4)
Δσ − 1 σ y Marin [22] deviates from the others and it cannot be approximated with
reasonable accuracy by Eqs. (2), (5) and (7). These relationships in
where the numerical subscript refers to the stress ratio (in this case R =
[16,20–22] make use of the tensile strength, however, some sources only
− 1). The equation strictly applies to the constant amplitude fatigue
give the nominal value or the measured yield stress.
limit but it is often considered as well for other stress ranges. The
The collected fatigue test data that will be presented in the following
resulting equation can be written in a more general form as:
sections are evaluated for the best fit values of parameters ξ,ζ, and η per
1− R type of detail. As an example, Fig. 2(a) provides the collected failed test
Δσ R = Δσ 0 ( ) (5)
data of DCC with preloaded high strength friction grip bolts, where the
1 + R Δζσσy0 − 1
gross section of the specimen is used to calculate the stress range. Three
of the series from [10] are displayed in colours. These are carried out at
where ζ = 1 for Soderberg and ζ = σu /σy in case of Goodman [16], R = − 0.6, R = 0 and R = 0.5 but the specimens are produced with
where σ u is the material tensile strength. identical material, surface and hole forming methods. Parameters ξ, ζ,
The Walker equation [17] assumes that the fatigue resistance is not and η are determined using these three series such, that the scatter of the

Table 1
FAT classes according to AASHTO:2012 [1], EN 1993-1-9:2005 [2] and BS 7608:2014 [3].

a)
FAT 125 for drilled or reamed holes, FAT 71 for punched holes.
b)
Shear stress in bolt shaft. Thread not in shear plane.
c)
FAT 97 for drilled or reamed holes, FAT 54 for thermally cut holes.
d)
Reduction of the FAT class for large diameter bolts in EN 1993-1-9 and BS 7608.

3
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

S-N curve is minimised. This gives ξ = 0.4,ζ = 1.1, and η = 0.6. These fatigue reference resistance, defined as the 95% lower prediction bound
corrections are subsequently applied to all data collected on that detail. of the fatigue resistance at 2 million cycles and at a stress ratio of R =
Fig. 2(b) provides the data for stress ranges corrected to a stress ratio of 0.5, is denoted as ΔσC :
R = 0 using Morrow’s Eq. (2) with ξ = 0.4. The figure shows a sub­ ( ( ) )
a − log 2⋅106 − kn s
loĝ
stantial reduction of the scatter as compared to the original data. A logΔσ C = (12)
similar result is obtained with Walker’s Eq. (7) and η = 0.6. The scatter m
̂
after stress ratio correction is larger in case of Soderberg’s Eq. (5) with where kn is the prediction bound factor:
ζ = 1.1. (Section 3.1 gives a further elaboration of this detail.) Similar √̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅
evaluations are performed for the other types of detail and in most cases, 1 1
Eqs. (2) and (7) give a similar remaining scatter and they outperform Eq. kn = t0.05 + +f (13)
n q
(5). For this reason, all data in the subsequent sections are corrected for
the stress ratio using Eq. (2). where t0.05 is the 5% coefficient of Student’s T distribution with DOF
The stress ratio in most practical civil engineering structures ranges equal to λ = n − 3 (considering the three estimated parameters ̂ a, m̂ and
from very low to very high. However, for very high stress ratios, such as ξ), q is the number of future samples, which is taken as 1, and f is a factor
R > 0.5, the static resistance is often decisive for the steel grades and considering the confidence of the prediction at the specific stress range
structure types described before. In addition, test data with R > 0.5 are ΔσC , which is related to the relative difference between that stress range
rare. For a safe fatigue design of general application, the fatigue resis­ and the centre – or average stress range – of the data [23]:
tance is therefore evaluated at a stress ratio of R = 0.5 (Δσ0.5 ) in the
( )2
subsequent sections. A separate evaluation is done for bolts loaded in loĝ
a− log(2⋅106 ) ∑
1
tension, Section 3.5. ̂
m
− n logΔσ0.5,i
f = ( )2 (14)
∑ ∑
logΔσ 0.5,i − 1n logΔσ0.5,i
2.3. Statistical evaluation of test data
Sub-groups are defined for most of the details regarding the pro­
An S-N curve is derived for each dataset. The Basquin relationship is duction method or geometry. The Kolmogorov-Smirnov test [24] is
assumed between N and Δσ0.5 : applied on the values of a per test to verify whether the subdivision into
logN = loga − mlogΔσ0.5 (8) subgroups is justified. This statistical test indeed rejected the null hy­
pothesis that the distinguished sub-groups belong to the same distribu­
where parameters m and loga are detail dependent and based on failed tion for every defined sub-group.
specimens only (i.e. excluding run-outs). They are derived for the 10- In pooling data of series with (remaining) differences in material
base-logarithm. It is assumed that loga is normally distributed. The es­ properties, surface smoothness, production process, geometry or stress
timates of m and a based on the test data consisting of n failed tests are ratio, the data fit with the Basquin relationship Eq. (8) may be worse
denoted m̂ and ̂ a , respectively: than in case of an individual series. The nature of the Basquin rela­
∑( ) ∑ ∑ tionship is such, that a worse fit results in a smaller value of m.
̂ For this
− n logΔσ0.5,i ⋅logNi + logΔσ0.5,i logNi
m
̂ = ∑( )2 ( ∑ )2 (9) reason, m̂ is estimated for individual series, but only if the series covers
n logΔσ0.5,i − logΔσ 0.5,i at least 1.5 decades of N as a smaller coverage may lead to an inaccurate
determination. The DOF increases to λ = n − 2 because of the absence of
1(∑ ∑ )
ξ as a parameter in case of an individual series. The fatigue reference
a=
̂ logNi + m
̂ logΔσ0.5,i (10)
n resistance Δσ C (Eq. 12) is determined for all series pooled per sub-group
The estimate of the standard deviation of loga is denoted with s: of detail type and using fixed (prior) slope parameters m = 3 and m = 5.
√̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅̅ The DOF is λ = n − 2 in that case because m is taken as prior.
∑( )2
logNi − loĝ a+m ̂ logΔσ 0.5,i All data in the subsequent sections are presented in graphs with a
s= (11) maximum stress range 20 times larger than the minimum stress range on
λ
the ordinate, thereby enabling a visual comparison between the graphs.
where λ is the Degree Of Freedom (DOF), quantified below. The

Fig. 1. Data from plates with cut holes at stress ratio R = 0.1 from [4]: (a) Oxy-fuel cut holes; (b) Plasma cut holes.

4
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

Fig. 2. Data for double covered connections (DCC) with preloaded bolts: (a) As-received; (b) Corrected to a stress ratio R = 0.

3. Test data per detail type maximum gross stress close to or even exceeding the yield stress. The
applied stress is distributed between the main plate and the covered
3.1. Double covered connections (DCC) with preloaded bolts plates, however, it can result into a high Von Mises stress if combined
with the compression stress in plate thickness direction caused by bolt
It is well known that the load transfer in DCC depends on whether the preloading. Yielding may influence the fatigue resistance particularly
bolts are preloaded [25]. The forces are transferred through friction for this detail, as it may alter the contact between the components. Fig. 3
between the plates in DCC with preloaded high strength friction grip presents the data again (with the gross stress range on the ordinate), but
bolts. The associated failure mode is then usually fretting fatigue at the including RunOut tests (RO) and with the data corrected to a stress ratio
end of the washer or sometimes at the end of the cover plate, as of R = 0.5. Three subgroups are distinguished:
demonstrated among others by [26,27]. The plate surface treatment and
preloading procedure should be subject to careful execution, or other­ • The maximum stress of the cycle exceeds the yield stress (red dots).
wise, the fatigue resistance may be significantly lower. As an example, • The maximum stress is between yield and 90% of yield (orange dots).
fatigue cracks starting from the hole edge were observed in [28] and the • The other data (black dots).
fatigue resistance was significantly lower than that of other series. This
was attributed to the oil used for drilling the holes, which has penetrated An additional series is added with a high yield stress (blue squares),
between the plates. Even in case of no apparent errors in the preloading which has a maximum ratio σR,max /σ y of 0.45. The data show a clear
procedure, cracks starting from the hole are occasionally reported. influence of the ratio σR,max /σ y . The higher this ratio, the flatter is the S-
Yin et al. [29] and Albrecht et al. [30] state that the fatigue resistance N curve. The data with σy ⩾827 MPa are on the lower bound of the entire
reduces as the number of bolt rows increases, because the frictional force pool for endurances greater than 2⋅105 cycles but they clearly exceed the
transfer near the first bolt row is larger than that of the others. A fatigue resistance of other series for lower endurances. The statistical evaluation
resistance equal to that of a hole in a plate is reported for four (or more) was performed without these high strength steel grades. The fatigue
rows of bolt per side of the connection. In contrast, tests in [31] with four reference resistance for an assumed slope of m = 5 increases from 81
bolt rows give a similar fatigue resistance as those with two bolt rows per MPa considering all (other) data to 84 MPa when excluding the data
side. The difference between these sources may be related to the safety with σ R,max /σy ⩾1 and to 89 MPa when excluding the data with
against slip. These connections are usually designed for slip resistance σ R,max /σy ⩾0.9. Excluding these subgroups has a negligible influence on ξ
under static loading, in which the resistance is determined as the sum­ and on ΔσC for m = 17, which gives Δσ C = 112 MPa.
med preload forces of all bolts multiplied by the slip factor. In the A few data at σmax /σy < 0.9 (near N = 105 and at N = 5⋅106 ) are
weakest link failure mode of fatigue, it is the first bolt row that is below the general trend. This may have been caused by unintended slip.
decisive regarding slip force and bypass loading. The slip requirement As this may also occur in practice, these outliers have been included in
may hence be stricter for fatigue as compared to the static failure mode the statistical evaluation.
in connections with a large number of bolt rows. The data in Fig. 3 include HDG steel series from [10]. They have been
Table A.1 provides the series used in the statistical evaluation. Data added because these data allow to evaluate ξ because of the large
are only considered for which the preload force and preload procedure coverage of stress ratios. The non-galvanized data were all carried out
are fully documented, in agreement with modern standards, and theo­ between 0⩽R⩽0.25. The HDG series at R = 0 in [10] gave an average
retically sufficient to prevent slip. The slope parameter of the individual fatigue resistance that was 6% larger than that of the non-galvanized
series range between 5.3⩽ m ̂ ≤ 51 with an average of m ̂ = 17. This is series of the same source and with the same stress ratio. Excluding the
substantially higher than the slope parameters of the other detail types HDG series (but keeping ξ) has no effect on the fatigue reference resis­
that will be discussed in subsequent sections. The data corrected to a tance of the entire database.
stress ratio of R = 0 were already presented in Fig. 2b. These data sug­ A special case is formed by DCC with preloaded injection bolts. Tests
gest that the S-N curve of the pooled data runs flatter at high stress have shown that the resin does not fail in fatigue [32]. However, a
ranges than at low stress ranges. Considering the data more closely, it (small) fraction of the load may be transferred through bearing in such a
appears that a significant number of the tests were carried out with the connection [32]. Only one source with fatigue tests was found on this

5
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

Fig. 3. Data for double covered connections (DCC) with preloaded high strength friction grip bolts, showing the influence of the maximum stress.

type [33], where it was observed that the fatigue resistance with pre­ stress ratio but also those with lower stress ratios. This difference cannot
loaded injection bolts was lower than that of normal preloaded high be explained by a simple stress ratio correction alone. The scatter is
strength friction grip bolts. believed to be (also) related to the differences in geometry. Therefore,
this paper proposes a new definition of the stress range for this detail
type, which is derived as follows: The geometry of the connection de­
3.2. Double covered connections (DCC) with non-preloaded bolts
termines which fraction of the load is passing the hole (bypass loading)
and which fraction is transferred through bearing with the bolt acting as
The forces are transferred via shear of the bolts and bearing of the byp
plates in DCC with non-preloaded bolts. Two potential failure modes a pin. The SCF-s for bypass and pin loading, Kt and Kpin t respectively,
apply in this case, namely, failure of the plate or failure of the bolt. In the are different (Fig. 5):
first failure mode of plate failure, the largest stress concentration occurs (
d0
) (
d0
)2 (
d0
)3
at the hole and this is the location where cracks initiate. For this reason, K byp
t = 2 + 0.284 1 − − 0.6 1 − + 1.32 1 − (15)
w w w
data are usually evaluated using the net section stress range Δσ R,net .
Valtinat and Huhn [10] have demonstrated that the fatigue resistance of d0
( )2
d0
( )3
d0
DCC with bolts with a small preload – not preventing slip – is larger than K pin
t = 12.88 − 52.71 + 89.76 − 51.67 (16)
w w w
that of DCC with bolts without any preload. Therefore, test data are only
included in the evaluation if the sources informed that the preload of the where d0 is the hole diameter and w is the plate width divided by the
bolts was negligible, such as in snug-tight bolts or hand-turned nuts. number of bolts over the width. The equations are taken from [34] and
Table A.2 and Fig. 4(a) show the collected data. The slope parame­ they apply to the net section stress.
ters of the individual series ranged between 4.3⩽ m⩽8.9,
̂ with an average The number of rows of bolts per side of the connection is denoted k.
value of 5.8. A large scatter results when pooling all data. The data with According to [35], the first row in connections with k > 1 transfers a
R = 0.3 generally give a higher endurance than the data with higher

Fig. 4. Data for double covered connections (DCC) with non-preloaded bolts in normal clearance holes: (a) Net section stress; (b) Modified net section stress at R =
0.5.

6
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

Table 3
Proposed FAT classes (constant amplitude fatigue limit at N = 2⋅106 ).
Detail m FAT Stress calculation

A. DCC with preloaded bolts 5 112 Gross section

B1. Plates in non-preloaded DCC, drilled or 5 90 Modified net section, Eq.
reamed holes (19)a)
B2. As B1, but punched or thermal cut 3 71 See B1
holes
B3. Bolts in non-preloaded DCC 5 100 Average τ per shear
planeb)
C. Single lap connection, preloaded bolts 5 100 Gross sectionc)
D1. Round holes drilled or reamed 5 90 Net section
D2. Round holes punched or thermal cut 3 50 Net section
E1. Bolt HT/R in tension 3 71 Tensile stress aread)
E2. Bolt R/HT in tension 3 56 See E1
E3. Bolt HDG or cut thread in tension 3 50 See E1

a) With c2 and c3 of Table 2. Use k = 1 for normal clearance holes.

b) Thread not in shear plane. Use bolts of grade 5.6 or higher.
c) Must be supported out of plane.
d) Multiply FAT with Eq. (20) with ν = 0.25 for bolts with D > 30 mm.

Table A.1
DCC with preloaded bolts.

Fig. 5. Stress concentrations in DCC with non-preloaded bolts: (a) Bypass Source σy [MPa] σu [MPa] k R n nRO
loading; (b) Pin loading.
[10]a) 290 430 1 − 0.6⩽R⩽0.5 89 22
[69]b) 280–470 ? 2 0.1 80 0
slightly larger fraction of the force than the other rows, whereas [29] [70] 286 ? 1, 2 or 3 0.5 6 0
informs that each row takes an equal fraction. The latter approximation [71] A588 ? 1 or 2 0 20 0
results into the following equation for the SCF at the first (decisive) bolt [72] 372 490 2 ≈ 0.1 1 5

row: [73] A514 689 4 ≈ 0.25 5 0

[37]c) 235–827 ? 2 or 4 0 62 0
1 k − 1 byp [31] 373 491 2 or 4 0 25 3
Kt = K pin + Kt (17)
k t k
a) Contains normal and HDG plates.
b) Contains one series of weathering steel.
Eq. (17) is simplified (with a coefficient of determination of 0.99 for the
c) Data of high strength steel (σy = 827 MPa) only used for comparison.
relevant range of 0.05⩽d0 /w⩽0.5) through:
( )3
d0
Kt ≈ c1 + c1 c2 − c3 (18) Table A.2
w
DCC with non-preloaded bolts in normal clearance holes, plate failure.
with coefficient c1 = 2.55 and coefficients c2 and c3 according to Source σy σu Holes k R n nRO
Table 2. The proposed stress range for evaluation of this detail type is: [MPa] [MPa]
( ( )3 ) [10]a) 290 430 Drilled or 1 0.1 or 96 14
Kt d0 punched 0.5
ΔσR,mod = Δσ R,net = Δσ R,net 1 + c2 − c3 (19)
c1 w [74] 420 475 Drilled 2 or 0.3 27 4
3
Using data from individual series and geometries, the stress ratio [75] 224 431 Drilled 1 0 1 5
effect was established as ξ = 0.5. Fig. 4(b) shows the data as a function [76] 283 454 Drilled 2 0.15 1 0
[72] 372 490 Drilled or 1 or ≈ 0.15 6 0
of the modified stress range corrected to stress ratio R = 0.5. The
punched 2
significantly reduced scatter of the test data compared to Fig. 4(a) in­
dicates that the newly defined stress range is an improvement over the a) Contains normal and HDG plates.
net section stress range.
The data in Fig. 4 include HDG steel samples from [36] because the with punched holes.
database excluding HDG steel is not large with 45 failed samples. The DCC with non-preloaded bolts in closely fitting holes – called fitted
HDG steel series gave the same fatigue resistance as those of non- bolts hereafter – are also evaluated using Eq. (19), because the SCF-s of
galvanized series with the same geometry and stress ratios in [36]. these detail types are similar. As the number of collected tests is small,
The fatigue reference resistance of the sub-group without HDG speci­ data of old double covered hot-riveted connections are added to the data
mens is 1% higher than that of the entire database. A sub-group con­ pool, but only those with red lead paint applied at the contact faces, as
taining specimen with drilled holes in the plates gives a 5% higher this results into a very low friction coefficient [37]. These riveted con­
fatigue reference resistance as compared to a sub-group of specimens nections are expected to perform similarly as DCC with fitted bolts
because of this low friction coefficient in combination with the generally
low clamping force of rivets [38].
Table 2 Table A.3 and Fig. 6 show the collected data. The slope parameter of
Coefficients c2 and c3 as a function of the number of bolt rows k. the series with fitted bolts is m
̂ = 4.1. The data have a stress ratio of
k 1 2 ⩾3 0⩽R⩽0.25. The same stress ratio correction is applied as for non-
preloaded bolts in normal clearance holes, i.e. ξ = 0.5. The fatigue
c2 1.6 1.3 1.1
resistance for the riveted connections is slightly below that of the fitted
c3 2.7 2.2 1.8
bolts. This may be related to heat treatment of the plate material during

7
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

Table A.3 conservative for bolts loaded in shear with high stress ratio. If applied so,
DCC with non-preloaded fitted bolts or with rivets and red lead paint, plate a conservative estimate of the fatigue reference resistance results of
failure. ΔσC = 98 MPa for m = 5.
Source σy [MPa] σu [MPa] Type R n nRO It should be noted that this value is derived for DCC. The fatigue
resistance may be lower in case of single lap connections because of the
[77] 332 517 Fitted Bolts 0.05 5 1
[78] 440 605 Rivets 0.14 4 0
bending moment introduced in the bolts, but fatigue test data with this
[79] 397–460 578–607 Rivets 0.11 8 2 condition are not found. The bolts of series [39] are of grade 5.8. Bolts of
[80] (mild steel) (mild steel) Rivets 0.15 10 0 higher grades are expected to have an equal or larger fatigue resistance.
Lower strength bolts may have a lower fatigue resistance, but data are
lacking. Lower strength bolts should hence not be used without sub­
the riveting process, but it may also be related to the uncertainty in
stantiation of the fatigue resistance by tests.
fatigue resistance associated with the small sample size. The number of
series is too small to conclude whether the modified stress is a better
3.3. Single lap connections with preloaded bolts
predictor than the nominal stress for connections with fitted bolts.
However, the fatigue resistance using the modified stress range is indeed
Single lap connections differ from DCC by having only one shear
almost equal to that of DCC with non-preloaded bolts in normal clear­
plane. Even when preloaded, these connections show a lower fatigue
ance holes in Fig. 4, which is in agreement with the theory of stress
resistance than DCC, which is attributed to the unsymmetrical force
concentration.
distribution [30,37]. Single lap connections are often applied in situa­
The second failure mode of DCC with non-preloaded bolts is fatigue
tions were only a part of the cross-section is connected. This results into
of the bolt loaded in shear. The stress range ΔτR is defined as the force
a geometry specific stress concentration [30], which should be deter­
range per shear plane divided either by the gross area of the bolt if the
mined on a case basis. Only data where the full cross-section is con­
shear plane is in the non-threaded part, or by the stress area if the shear
nected are used in the statistical evaluation.
plane is in the threaded part. Only two sources are available, with data
Table A.5 and Fig. 8 present the data. The average values are m ̂ = 3.7
given in Table A.4 and Fig. 7. The figure demonstrates a very large in­
and ξ = 0.6. Approximately 50% of the data stem from [44], where the
fluence of the shear plane location. For this reason, bolts with a shear
steel applied for plate and section material has been produced with the
plane through the threaded part should be avoided in such connections
Thomas process. The possibly low fracture toughness of this type of steel
and the fatigue reference resistance in Fig. 7 is evaluated for the other
may have influenced the fatigue performance. Although the Thomas
data, i.e. source [39]. The slope parameter of the data in [39] is m̂ = 8.
process is no longer used, the data are still included because of
As only one series is available, the resulting fatigue resistance is
demonstration purposes and because the number of other data is small.
compared to that of riveted connections failing through rivet shear in
Fig. 8 shows a large scatter, even when considering only data from [44]
[40,41]. The resulting rivet fatigue resistance assuming m = 5 is ΔσC =
for R = 0. One of the reasons for the large scatter is the difference in the
140 MPa. This result is in line with that of the bolts: Δσ C = 135 MPa for
support against out of plane deformation. To demonstrate this, Fig. 8
the same slope.
highlights two series from [44]:
All collected tests are carried out at relatively low stress ratios of
R ≈ 0.1. Davoli et al. [42] showed that the torsional fatigue resistance
depends only slightly on the mean shear stress as long as the maximum
shear stress does not exceed the static shear yield stress. Bennebach et al.
[43] showed that the effect of the mean shear stress is larger in case of Table A.4
block loading with varying mean shear stress. However, both studies DCC with non-preloaded bolts, bolt failure.
apply to a steel of much higher grade than those considered here. Most of
Source Grade shear plane R n nRO
the data of the other details, where components are loaded in tension,
show a stress ratio effect that is on average ξ ≈ 0.55. In the light of the [39] 5.8 plain 0.1 13 2
references mentioned above, this stress ratio effect is probably [31] 8.8–12.9 threaded 0.1⩽R⩽0.35 28 2

Fig. 6. Data for double covered connections (DCC) with fitted bolts or rivets: (a) Net section stress; (b) Modified net section stress at R = 0.5.

8
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

Fig. 7. Data for double covered connections (DCC) with bolt failure.

is performed only for series 2, giving a fatigue resistance of ΔσC =

Table A.5
104 MPa for m = 5.
Single lap connections with preloaded bolts.
Source Grade R n nRO The other series have a lateral support in between these two extreme
[81] ASTM-A7 − 1⩽R⩽0 30 0 cases and they have a fatigue resistance that is also in between, Fig. 8(b).
[44]a) Thomas − 1⩽R⩽0.1 36 4 Also for this type of connection [33] found that the fatigue resistance
[33]b) Fe510 0.1 3 0 with preloaded injection bolts was lower than that of normal preloaded
bolts.
a) Only data with all connections fully supported are considered.
b) Data from puddle iron specimens excluded.
3.4. Plates with holes
• Series 1 has no lateral support. The distance between the grips in the
test machine relative to the specimen cross-section was large. The Many tests series are found for plates with holes, see Table A.6.
fatigue resistance of this series is lower than that of the others. Several sources such as [45] show that the fatigue resistance of plates
• Series 2 (actually consisting of two geometries) has a full lateral with slotted holes is different from that of plates with round holes, in line
support through the webs of the sections. The fatigue resistance of with the expectation regarding the theoretical SCF. Only round holes are
this series is higher than that of the others. The statistical evaluation therefore considered. The fatigue resistance of plates with holes depends
on the hole forming method, see e.g. [45,46]. Three methods are

Fig. 8. Data for single lap connections with preloaded bolts: (a) As received; (b) Corrected to R = 0.5.

9
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

Table A.6 other data. This may be caused by local yielding at the hole edge at
Elements with a round hole. maximum stress in the first cycle(s) in the tests with high stress ratio.
Source σy [MPa] σu [MPa] Hole forming R n nRO This has no effect on the stress range, but the effective stress ratio at the
hole edge is then lower than the ratio of the externally applied load in
[82]a 250–345b ? various 0.1 28 8
the tests with high stress ratio. Hence, it may be conservative to correct
[83] 235–355b) ? drilled 0.1 17 4
[84]c 235–355b) ? drilled 0.1 16 3 the data for 0⩽R⩽0.1 to that of R = 0.5. The average slope parameter of
[85] 235–390 396–641 drilled or 0 47 8 the individual series is m̂ = 6.6. Using the net section stress range and
reamed assuming a slope parameter of m = 5, the fatigue reference resistance is
[4] 427–484 559–596 thermal cut 0.1 65 25 97 MPa.
[86] 359–538 511–596 drilled or 0.1 54 13
punched
Only data with 0⩽R⩽0.1 have been found for plates with thermally
[79] 248–422 400–607 drilled 0⩽R⩽0.28 68 33 cut holes (except for one run-out at R = 0.5) and punched holes. The
[87] 386 ? drilled − 1⩽R⩽0.5 29 9 data were corrected for the stress ratio by assuming the same correction
[10] 235b) ? drilled or 0.1 38 19 factor as for drilled holes, i.e. ξ = 0.65, Fig. 10. The slope parameters of
punched the individual series are m ̂ = 3.7 and 3.5 for thermally cut holes and
[88] 229–643 408–692 reamed 0 23 8 punched holes, respectively. The steeper slopes and lower fatigue
[89] 355–460b) ? drilled or − 1⩽R⩽0.1 134 0
reference resistances of thermally cut or punched holes relative to
reamed
[45] 355b) ? drilled or oxy- 0.1 45 2 drilled or reamed holes indicates the appropriateness of the subdivision
fuel into these groups.
[90] 355b) ? drilled − 1⩽R⩽0 25 3 A large range of plate and hole dimensions can be applied in practice,
i.e. the variation in relative dimensions is larger than in case of the
a) Data from HDG excluded.
b) Nominal yield stress (no other data given). connections in the previous sections. The theoretical SCF, Eq. (15),
c) Staggered holes. suggests an influence of ratio d0 /w. The SCF further depends on the hole
diameter over plate thickness ratio, evaluated in [49,50]. A fatigue
notch factor can be derived from the theoretical SCF by considering the
distinguished:
stress gradient near the notch according to [7]. Filippini [51] provides
data for the stress gradient in plates with holes. It appears predomi­
• Drilled holes, or holes that are reamed after any of the following two
methods. Cracks initiate at the hole, near the centre of the plate nantly related to the ratio d20 /w. The fatigue notch factor of the test
thickness or at the plate surface for thick or thin plates, respectively. specimens is estimated from the equations or digitized figures in these
• Thermally cut holes (Oxy-fuel, plasma or laser). These three methods sources. It is 2.16 on average, with a standard deviation of 0.08. Because
appear to give a similar fatigue performance. The Vickers hardness of the low standard deviation, i.e. the similar geometries of the speci­
was generally below 400 HV5. mens, an evaluation using the notch stress per specimen did not give a
• Punched holes. Cracks initiate at the hole at the plate surface [47]. noticeable lower scatter of the data, nor a different value for the fatigue
reference resistance. The net section stress without modifications is
Plates with waterjet cut holes as considered in [48] may have a higher therefore used in the evaluation, as this simplifies the evaluation for
fatigue resistance as compared to the three distinguished groups, but practice. BS 7608:2014 [3] provides an upper bound for the fatigue
they are not considered here because waterjet cutting is seldomly notch factor of 2.4 that should cover the majority of plate and hole
applied in civil engineering structures. Data with different stress ratios dimension combinations in practice. The fatigue resistance may in such
were found for plates with drilled or reamed holes, Fig. 9(a), but most cases hence be about 10% lower as compared to the values in Figs. 9 and
data are in the range 0⩽R⩽0.1. A best fit of individual series is obtained 10.
by correcting for the stress ratio with ξ = 0.65. Fig. 9(b) shows that such
a correction means that the data for 0⩽R⩽0.1 are on average below the

Fig. 9. Data for plates with drilled or reamed holes: (a) As received; (b) Corrected to R = 0.5.

10
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

Fig. 10. Data for plates with holes, corrected to R = 0.5: (a) Thermally cut; (b) Punched.

3.5. Bolts loaded in tension

Bolts loaded in tension usually fail at the first thread in the nut,
sometimes at the neck of the bolt head, and incidentally at the first
thread adjacent to the unthreaded shank. The shape of the thread de­
termines the stress concentration of the dominant failure mode. One of
the consequences is that bolts with fine thread have a different fatigue
resistance as compared to that of bolts with coarse thread. Civil engi­
neering structures make almost exclusively use of bolts with ISO metric
coarse thread [52]. For this reason, only bolts with that thread are
evaluated.
Bolts loaded in tension are usually preloaded, as this significantly
reduces the force range in the bolt. The fraction of the force range
transferred through the bolt depends on the stiffness and composition of
the plate assembly. Plates should be carefully leveled, as unintended
prying effects due to imperfect plate assemblies may increase the force
fraction in the bolt considerably [53]. Imperfections of plates are not
considered in the evaluation below.
The nominal preload stress is usually equivalent to 0.7σ u . Table A.7 Fig. 11. Data for bolts in tension.
and Fig. 11 present the data on bolts loaded in tension, where a
distinction is made between specimens with a high and with a low mean attributed to effects of imperfections. Strain measurements on bolts with
stress. The stress in this and in subsequent figures is defined as the force a low preload in [31] demonstrate that the stress range is not equal
divided by the stress area of the bolt. The data of bolts with a low mean around the perimeter of the bolt if that bolt is slightly croocked. Only the
stress give on average a longer but also a more scattered fatigue life as data with a mean stress > 0.6σ u are considered hereafter.
compared to bolts with a high mean stress. The larger scatter is Almost all specimens are of grade 8.8 or 10.9. The fatigue resistance
of these two grades appears similar. The few specimens of lower grade
suggest that lower grades give a slightly better fatigue performance. This
Table A.7 is attributed to the lower mean stress in these bolts when preloaded to
Bolts in tension. 70% of the tensile strength.
Source Grade Type D σ m /σ u n nRO Subgroups are selected based on the following production methods,
[mm] as they appear to influence the fatigue resistance:
[67] 8.8 R/HT 14 0.7 27 10
[91] 10.9 HDG 36 0.73 23 8 • Heat treated and then hot-rolled bolts (HT/R). Such bolts may
[92] 10.9 unknown 20 0.7 37 0 benefit from the compressive residual stresses caused by rolling at
[57] 10.9, 12.9 HT/R 36–72 0.7 43 27 the crack initiation locations in the thread.
[31] 8.8, 12.9 HT/R, R/HT, 12–36 0.25–0.55 57 4
Cut
• Rolled and then heat treated bolts (R/HT).
[93] 10.9 R/HT 20 0.25–0.95 41 0 • One of the previous methods, followed or preceded by Hot-Dip
[94] 10.9 HDG 48 0.15 27 4 Galvanizing (HDG). Micro-cracking in the zinc layer may nega­
[95] 10.9 R/HT, HDG 36 0.7 50 17 tively affect the fatigue resistance [54].
[54] 10.9 R/HT, HDG 64 0.7 17 1
• Bolts with cut thread (Cut). The potentially sharp thread geometry
[64] A36 A193 HT/R, cut 35–51 0.3–0.6 32 4
4340 compared to rolling may negatively affect the fatigue resistance [55].
[96] 10.9 R/HT 20 0.65 29 0
[97] 10.9 unknown 36 0.1–0.85 15 0 One of the consequence of the ISO course pitch metric thread is that the
a)
10.9 R/HT, HDG 12–36 0.63 46 5 fatigue resistance depends on the bolt diameter. Different standards give
a) Personal correspondence with Fraunhofer LBF, Germany. different values for the bolt diameter influence. AASHTO [1] does not

11
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

specify a diameter influence. EN 1993-1-9 [2] and BS 6708 [3] give the between the evaluations using the mean resistance (open circles) or
following reduction on the fatigue resistance for bolt diameters D larger using the 95% exceedance fraction (filled circles) of that source dem­
than 30 mm: onstrates that the evaluation of the diameter effect is sensitive to the
( )ν scatter in fatigue test data.
30 mm
ΔσC = ΔσC,M30 (20) The thickness effect is also evaluated based on all fatigue test data, by
D
assuming Eq. (20) for all bolt diameters (i.e. also for D <30 mm) and
fitting ν such that s is minimized. Only results where failure occurred at
with ν = 0.25. The guidelines in VDI 2230 [56] provide a smaller in­
N⩽106 cycles are considered in the fit, because the Basquin relation of
fluence of the bolt diameter as compared to this equation. There is a
Eq. (8) does not describe the data well for larger endurances. A best fit is
growing interest from practice in large diameter bolts. Three sources of
obtained for ν = 0.5, 0.15 and 0.05 for the subgroups of HT/R, R/HT and
information are used here to evaluate the influence of the bolt diameter:
HDG, respectively, but the evaluation is relatively insensitive to ν. It
should be mentioned that more data are available for the R/HT and HDG
• Tests on HT/R bolts in [57] were carried out with bolt diameters
subgroups than for the HT/R subgroup. Because of this insensitivity and
ranging between 36 mm ⩽ D⩽72 mm. Assuming a slope parameter of
because of the different results for the different evaluation methods, the
m = 3, the mean fatigue resistance at 2 million cycles is determined
data are evaluated using the reduction according to EN 1993-1-9, which
for each bolt diameter. These are then compared. Tests on M36 and
is expected to be generally conservative. Fig. 13 shows the result of this
M64 R/HT bolts are evaluated in a similar way in [58]. These data
evaluation. Only few data are available for bolts with cut threads. Data
are also considered.
of cut thread with σm /σu < 0.65 confirm the general trend of the S-N
• The theoretical SCF using nominal thread geometry is determined
curve and are all above the data shown.
with the Finite Element Method (FEM). An axi-symmetrical model is
Fig. 13 show a large difference in fatigue resistance between some of
made in software Abaqus v62.0, see Fig. 12(a) for the geometry and
the subgroups, a relatively small standard deviation per subgroup, and a
the mesh. Linear elements of type CAX4 are used for the elements
long transition from the finite to the near infinite life region. This
and frictional contact is applied between the elements with a friction
gradual transition is also evident from the data on other details, but less
coefficient of 0.1, considered representative for lubricated bolts [59].
pronounced as for bolts. The Basquin relation Eq. (8) does not describe
Such a model is not able to determine the exact value of the stress
the transition. For this reason, the data are also fitted with the Six
concentration because the gradual introduction of the first thread in
Parameter Random Fatigue Limit Model (6PRFLM) [62], which is a
the nut cannot be considered in such a model. However, it is ex­
random fatigue limit model based on [63]. Here, use is made of all data,
pected that this method allows to compare SCF-s of different bolt
diameters relative to each other. The SCF and the stress gradient including RO and failures at N > 106 cycles. First step in the 6PRFLM is
resulting from the model are used to determine the fatigue notch to estimate the fatigue limit Δσlim , which requires a sufficient number of
factor, Kf , using [7]. run-outs at larger number of cycles than the failed data. For this reason,
• A crack growth calculation using a fracture mechanics (FM) model is only data on HT/R and HDG bolts are fitted to the 6PRFLM. The 6PRFLM
carried out. Equations from [60] are used for the stress intensity S-N curve reads:
factors, which are checked with results in [61]. A Paris equation with ( Δσ lim )
(21)
′ ′

an exponent of 3 is used to determine the crack growth rate. An logN = loga − m logΔσ − plog 1 −
Δσ
initial semi-circular (production) defect with a radius of 0.15 mm is
assumed and the calculation is terminated after a crack depth of 0.6 where the transition radius of the curve is controlled through parameter
times the bolt diameter is obtained. The calculation is used to esti­ p. Parameters a and m have the same purpose as a and m in the Basquin
′ ′

mate the ratios of the fatigue resistance considering crack growth for relationship, but with different values. Leonetti et al. [62] explain how
different diameters (hence the value of the Paris constant is not to obtain the distributions and the correlations of the parameters of the
important). 6PRFLM. The 95% prediction bound is determined with Monte Carlo
analysis using these distributions and correlations and assuming kn =
Fig. 12(b) presents the results in terms of the predicted fatigue 1.64. It is plotted with continuous curves in Fig. 13. The figure shows
resistance relative to that of an M36 bolt. The fatigue notch factor and that the slope parameter m is close to 3 and that the 95% prediction
′

the FM calculations agree well with the test data of R/HT bolts [58] and bound of Δσlim is close to the value of ΔσC of the Basquin relation for m =
these are in between VDI [56] and EN 1993-1-9 [2]. The data of HT/R 3.
bolts in [57], however, are generally below these models. The difference Threaded rods or stud bolts are not considered in the evaluation.

Fig. 12. Influence of bolt diameter on fatigue resistance: (a) Finite element model to determine the SCF; (b) fatigue resistance relative to that of an M36 bolt.

12
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

Fig. 13. Data for bolts in tension, σ m /σ u ⩾0.65, with diameter correction: (a) Heat treated then rolled (HT/R); (b) Rolled then heat treated (R/HT); (c) Galvanized
(HDG); (d) Cut threat (Cut).

Data from American sources [64–66] on these elements show that the parameter per detail is selected based on this generality as either m =
fatigue resistance is similar to that for bolts as presented above. It should 3 or m = 5.
be mentioned that the fatigue resistance of bolts or threaded rods loaded • Plates with drilled or reamed holes gave a higher fatigue resistance
in bending is substantially larger than the tension case discussed here and a higher slope parameter than plates with thermally cut or
[67]. punched holes. Although a smaller difference between these hole
forming methods was observed for net section failure in DCC with
4. Proposed FAT classes for the design non-preloaded bolts, the same distinction in hole forming method is
made for these connections, as a conservative approximation.
This section uses the evaluations of the previous section to propose • Assuming a slope m = 5, Section 3.1 shows that the fatigue resis­
FAT classes for the design of civil engineering structures. The lowest tance of DCC with preloaded bolts increases when omitting tests
predefined FAT class in EN 1993-1-9 is 36 and each subsequent FAT carried out with very high maximum stress and it approaches (but
class has a 12.5% higher fatigue resistance, rounded to integers. The FAT does not reach) the resistance associated to a free slope (ΔσC = 112
class is determined per detail as the fatigue reference resistance rounded MPa). Given the relevancy in practice of small stress ranges and high
down to the nearest predefined FAT class. endurances, it is therefore reasonable to adopt FAT class 112.
As explained, the fatigue reference resistance values are derived for • Sections 3.1 and 3.3 suggest that plate failure in connections with
steel grades with a nominal yield stress up to 460 MPa. Test data on preloaded injection bolts give a lower fatigue resistance than con­
higher strength details appear to generally give equal or higher fatigue nections with preloaded high strength friction grip bolts (without
endurances. The FAT classes derived can therefore be conservatively injection). However, this is based on only one source. The FAT class
applied to higher grade steels, within the general bounds set by the of connections with preloaded injection bolts is therefore not yet
Eurocode (grades up to S960). established.
A few judgement choices have been made: • Imperfections in the alignment of holes in the different plates of a
DCC with bolts may occur. If not preloaded, this may imply that not
• Generally, details with a longer fatigue initiation life and a higher all bolts participate in transferring the fatigue force. For this reason,
fatigue resistance have a larger slope parameter. This is confirmed by it is suggested that only one bolt row (k = 1) is considered in the
the slope estimate per series in the previous section. The slope calculation of the stress with Eq. (19). Such a situation is less likely

13
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

for fitted bolts, as the holes are then made in the entire plate as­ Acknowledgements
sembly at once.
• Rounding down to the nearest FAT class implies for plates with holes The members of the experts committees ECCS-TC6 and CEN/TC250/
that some allowance is available for the difference in fatigue notch SC3/WG9 are acknowledged for the discussions about the evaluation.
factor between the tests and the practical upper bound, Section 3.4. Helen Bartsch (RWTH Aachen), Sjors van Es (TNO) and Richard Pijpers
• Failures are observed at very large numbers of cycles, but a gradual (TNO) are acknowledged for collecting a number of sources with fatigue
transition is observed between the finite and the near infinite re­ test data. Matthijs Bakker (TNO) is acknowledged for producing the
gions. Using the Basquin relation with the 95% prediction bound, no finite element model of bolts. This work has not received funding.
single failure was observed below ΔσC (at two million cycles) and
only very few below the Basquin relation at one million cycles. The Appendix A. Description of the test series
6PRFLM applied to bolts confirmed that the near infinite life co­
incides with Δσ C . This applies to plates and bolts loaded in tension. This appendix gives the description of the test series and the used
Tests in [68] on smooth and scratched specimens loaded in shear also sources per detail type.
show a transition at about one or two million cycles. For design
practice, a constant amplitude fatigue limit may thus be assumed at References
two million cycles.
[1] AASHTO:2012. AASHTO LRFD Bridge Design Specifications. American Association
of State Highway and Transportation Officials; 2012.
The authors do not claim to have accounted for all influencing factors on [2] EN 1993-1-9:2005. Eurocode 3: Design of steel structures - Part 1-9: Fatigue. CEN;
the fatigue resistance per detail type. However, it is expected that the 2005.
dominant influencing factors are considered, which allows for an [3] BS 7608:2014. Guide to fatigue design and assessment of steel products. BSI; 2014.
[4] Cicero S, García T, Álvarez JA, Martín-Meizoso A, Aldazabal J, Bannister A, et al.
appropriate selection of the FAT class per detail. Table 3 gives these Definition and validation of eurocode 3 fat classes for structural steels containing
recommended FAT classes. These have been implemented in the revised oxy-fuel, plasma and laser cut holes. Int J Fatigue 2016;87:50–8.
version of EN 1993-1-9 that is sent to the European member states for [5] Cicero S, García T, Álvarez JA, Bannister A, Klimpel A, Martín-Meizoso A, et al.
Fatigue behaviour of structural steels with oxy-fuel, plasma and laser cut straight
commentary. edges. definition of eurocode 3 fat classes. Eng Struct 2016;111:152–61.
[6] Gurney TR. Fatigue of welded structures. Cambridge University Press; 1979.
5. Conclusions [7] FKM:2001. Rechnerischer Festigkeitsnachweis für Maschinenbauteile. VDMA
Verlag GmbH; 2012.
[8] Ferraz G, Rossi B. On the fatigue behaviour of hot dip galvanized structural steel
The evaluation of a large number of fatigue tests on bolts and bolted details. Eng Fail Anal 2020;118:104834.
connections with production qualities and dimensions relevant for civil [9] Bergengren Y, Melander A. An experimental and theoretical study of the fatigue
engineering structures have revealed the possibility to improve current properties of hot-dip galvanized high strength sheet steel. Int J Fatigue 1992;14:
154–62.
design specifications. These have been implemented in the revision of [10] Valtinat G, Huhn H. Betriebsfestigkeit von stählernen Lochstäben und
the European standard EN 1993-1-9. Schraubenverbindungen mit feuerverzinkten Bauteilen und gestanzten Löchern.
Düsseldorf: Gemeinschaftsausschuß Verzinken; 2000.
[11] Nilsson T, Engberg G, Trogen H. Fatigue properties of hot-dip galvanized steels.
• Most details benefit from sub-grouping into separate FAT classes Scand J Metall 1989;18:166–75.
based on the production process, such as the hole forming method in [12] Vogt JB, Boussac O, Foct J. Prediction of fatigue resistance of a hot-dip galvanized
plates or the thread forming method of bolts. steel. Fatigue Fract Eng Mater Struct 2000;23:33–9.
[13] Mischke CR. Prediction of stochastic endurance strength. J Vib Acoust Stress Reliab
• Most details show a gradual transition from the finite life to the near 1987;109:113–22.
infinite life regions of the S-N curve. The Basquin relation is therefore [14] DB-Richtlinie 805:2002. Tragsicherheit bestehender Eisenbahnbrücken. Deutsche
less suited to describe the data, but it is still proposed for the new Bahn; 2002.
[15] Soderberg C. Factor of safety and working stress. Trans ASME 1939;52:13–28.
generation of EN 1993-1-9 because of its ease in use. Using the [16] Goodman J. Mechanics applied to engineering. first ed. London: Longmans Green
Basquin relation, the near infinite life region commences at and Co; 1899.
approximately one or two million cycles for the details considered. [17] Walker K. The effect of stress ratio during crack propagation and fatigue for 2024-
t3 and 7075-t6 aluminum. In: Effects of Environment and Complex Load History on
• A substantial influence of the stress ratio, R, is observed. Expressing
Fatigue Life, ASTM STP 462. American Society for Testing and Materials; 1970. p.
the fatigue resistance at stress ratio R relative to that at R = 0 with 1–14.
the factor (1 − R)/(1 − ξR), the values of ξ established per detail and [18] Smith KN, Watson P, Topper TH. Stress-strain function for the fatigue of metals.
range between 0.4⩽ξ ≤ 0.65 for the fatigue test data ranging be­ J Mater 1970;5:767–78.
[19] Langlais TE, Vogel JH. Overcoming limitations of the conventional strain-life
tween − 1⩽R ≤ 0.5. The FAT classes are derived for R = 0.5. fatigue damage model. J Eng Mater Technol 1995;117:103–8.
• Most fatigue tests are carried out at 0⩽R⩽0.1. It is recommended to [20] Gerber W. Bestimmung der zulässigen Spannungen in Eisen-Constructionen.
carry out more tests at higher stress ratios to substantiate the stress Z Bayer Archit Ing Ver 1874;6:101–10.
[21] Dietmann H. Festigkeitsberechnung bei mehrachsiger Schwingbeanspruchung.
ratio influence. More data would also be welcome to establish the Konstruktion 1973;25:181–9.
size effect of bolts that are heat treated and then rolled. [22] Marin J. Interpretation of fatigue strength for combined stresses. In: Proceedings of
• Yielding has an important influence on DCC with preloaded bolts. international conference on fatigue of metals, institution of mechanical engineers;
1956. p. 184–94.
The FAT class should be based on data where yielding has not taken [23] Schneider CRA, Maddox SJ. Best practice guide on statistical analysis of fatigue
place. data. TWI doc. 13604.01/02/1157.02. The Welding Institute; 2002.
• The stress calculation in Double Covered Connections (DCC) with [24] Stephens MA. Edf statistics for goodness of fit and some comparisons. J Am Stat
Assoc 1973;69:730–7.
non-preloaded bolts should be based on the stress concentration at [25] Jimenez-Pena C, Talemi RH, Rossi B, Debruyne D. Investigations on the fretting
the hole edge. The net section stress is not appropriate for this detail. fatigue failure mechanism of bolted joints in high strength steel subjected to
A simple design equation is proposed. different levels of pretension. Tribol Int 2016;108:128–40.
[26] Juoksukangas J, Lehtovaara A, Mantyla A. Experimental and numerical
investigation of fretting fatigue behaviour in bolted joints. Tribol Int 2016;103:
Declaration of Competing Interest 440–8.
[27] Zampieri P, Curtarello A, Maiorana E, Pellegrino C. A review of the fatigue strength
The authors declare that they have no known competing financial of shear bolted connections. Int J Steel Struct 2019;19:1084–98.
[28] Wattar F, Albrecht P, Sahli AH. End-bolted cover plates. J Struct Eng 1985;111:
interests or personal relationships that could have appeared to influence 1235–49.
the work reported in this paper. [29] Yin WS, Fang QH, Wang SX, Wang XH. Fatigue strength of high strength bolted
joints. In: Proceedings of 3th IABSE congress. IABSE; 1982. p. 707–14.

14
J. Maljaars and M. Euler International Journal of Fatigue 151 (2021) 106355

[30] Albrecht P, Mecklenburg MF, Evans BM. Fatigue strength of bolted joints. J Struct [64] Fischer F, Frank K. Axial tension fatigue strength of anchor bolts. University of
Eng 1987;113:1834–49. Texas; 1977.
[31] [no names], Fatigue tests on steel bolts, report OTO 97067. Health and Safety [65] Frank K. Fatigue strength of anchor bolts. J Struct Div 1980;106:1279–93.
Executive; 1998. [66] Van Dien JP, Kaczinski MR, Dexter RJ. Fatigue testing of anchor bolts. In:
[32] Bouwman LP, Gresnigt AM. Design of connections with injection bolts. Delft Proceedings of the 14th Structures congress. ASCE; 1996. p. 337–44.
University of Technology; 1989. [67] Wentzel H, Huang X. Experimental characterization of the bending fatigue strength
[33] Correia JAFO, Pedrosa BAS, Raposo PC, De Jesus AMP, Santos Gervásio HM, of threaded fasteners. Int J Fatigue 2015;72:102–8.
Lesiuk GS, et al. Fatigue strength evaluation of resin-injected bolted connections [68] Nishimura Y, Yanase K, Tanaka Y, Miyamoto N, Miyakawa S, Endo M. Effects of
using statistical analysis. Engineering 2017;3:795–805. mean shear stress on the torsional fatigue strength of a spring steel with small
[34] Pilkey WD, Pilkey DF. Peterson’s Stress Concentration Factors. 3rd ed. Wiley; 2008. scratches. Int J Damage Mech 2020;29:4–18.
[35] Martin ST. Load distribution in bolted joints. Lehigh Universit 1960. [69] Lieurade HP. Etude de la tenue á la fatigue des assemblages boulonnés en aciers á
[36] Valtinat G, Huhn H. Festigkeitssteigerung von Schraubenverbindungen bei haute limite d‘élasticité. IRSID; 1976.
ermüdungsbeanspruchten, feuerverzinkten Stahlkonstruktionen. Stahlbau 2003; [70] Mas E, Janss J. Assemblages par boulons á haute résistance - Fatigue des
72:715–24. assemblages á double couvre-joint. CRIF; 1964.
[37] Kulak GL, Fisher JW, Struik JHA. Guide to Design Criteria for Bolted and Riveted [71] Albrecht P, Mecklenburg MF, Evans BM. Screening of structural adhesives for
joints. 2nd ed. American Institute of Steel Construction (AISC); 2001. application to steel bridges. US Department of Transportation, Federal Highway
[38] Leonetti D, Maljaars J, Pasquarelli G, Brando G. Rivet clamping force of as-built Administration; 1985.
hot-riveted connections in steel bridges. J Constr Steel Res 2020;167:105955. [72] Brown JD. Punched Holes in Structural Connections, Dissertation at The University
[39] Wichtowski B. Untersuchungen zur Ermüdungsfestigkeit von Paßschrauben beim of Texas, Austin; 2006.
Abscheren. Stahlbau 2007;76:235–40. [73] Frank KH, Yura JA. An experimental study of bolted shear connections. Austin:
[40] Brühwiler E, Hirt MA. Das Ermüdungsverhalten genieteter Brückenbauteile. University of Texas; 1981.
Stahlbau 1987;56:1–8. [74] Josi G, Grondin GY, Kulak GL. Fatigue of Bearing-type Shear Splice. University of
[41] Pipinato A, Pellegrino C, Bursi OS, Modena C. High-cycle fatigue behavior of Alberta, Department of Civil and Environmental Engineering; 1999.
riveted connections for railway metal bridges. J Constr Steel Res 2009;65:2167–75. [75] Wilson WM, Thomas FP. Fatigue tests of riveted joints. University of Illinois; 1938.
[42] Davoli P, Bernasconi A, Filippini M, Foletti S, Papadopoulos IV. Independence of [76] Steinhardt O, Möhler K. Versuche zur Anwendung vorgespannter Schrauben im
the torsional fatigue limit upon a mean shear stress. Int J Fatigue 2003;25:471–80. Stahlbau – I. Teil, Berichte des Deutschen Ausschusses für. Stahlbau 1954;18.
[43] Bennebach M, Palin-Luc T, Messager A. Effect of mean shear stress on the fatigue [77] Graf O. Versuche mit Schraubenverbindungen, Berichte des Deutschen Ausschusses
strength of notched components under multiaxial stress state. Procedia Eng 2018; für. Stahlbau 1951;16:1–19.
213:25–35. [78] Graf O. Versuche im Stahlbau - Dauerversuche mit Nietverbindungen. Berichte des
[44] Klöppel K, Seeger T. Dauerversuche mit einschnittigen HV-Verbindungen aus St 37. Deutschen Ausschusses für Stahlbau 1935;5:1–61.
Stahlbau 1964;33:225–45. [79] Klöppel K. Gemeinschaftversuche zur Bestimmung der Schwellzugfestigkeit voller,
[45] Mang F, Klingler J, Bucak O. Untersuchung der Dauerfestigkeit von Proben aus gelochter und genieteter Stäbe aus St37 und St52. Bautechnik 1936;13(14):
Stahl mit brenngeschnittenen Löchern. Schweißen und Schneiden 1989;41:81–5. 96–112.
[46] Sanchez L, Gutierrez-Solana F, Pesquera D. Fatigue behaviour of punched [80] [no names], Statistische Auswertung von Ermüdungsversuchen an
structural plates. Eng Fail Anal 2004;11:751–64. Nietverbindungen in Flussstahl. Office de Recherches et d’Essais. Union
[47] Alegre JM, Gutiérrez-Solana F, Aragón A. A finite element simulation methodology Internationale des Chemins de fer (ORE UIC); 1986.
of the fatigue behaviour of punched and drilled plate components. Eng Fail Anal [81] Munse W, Wright D, Newmark N. Laboratory tests of high tensile bolted structural
2004;11:737–50. joints. Trans ASCE 1954;80:1.
[48] Caiza P, Ummenhofer T. A probabilistic stüssi function for modelling the s-n curves [82] Brown JD, Lubitz DJ, Cekov YC, Frank KH, Keating PB. Evaluation of Influence of
and its application on specimens made of steel s355j2+n. Int J Fatigue 2018;117: Hole Making upon the Performance of Structural Steel Plates and Connections.
121–34. University of Texas; 2007.
[49] Yu P, Guo W, She C, Zhao J. The influence of poisson’s ratio on thickness- [83] Steinhardt O, Möhler K. Versuche zur Anwendung vorgespannter Schrauben im
dependent stress concentration at elliptic holes in elastic plates. Int J Fatigue 2008; Stahlbau – II. Teil, Berichte des Deutschen Ausschusses für. Stahlbau 1959;22.
30:165–71. [84] Steinhardt O, Möhler K. Versuche zur Anwendung vorgespannter Schrauben im
[50] Yang Z, Kim CB, Cho C, Beom HG. The concentration of stress and strain in finite Stahlbau – III. Teil, Berichte des Deutschen Ausschusses für. Stahlbau 1962;22.
thickness elastic plate containing a circular hole. Int J Solids Struct 2008;45: [85] Hansen NG. Fatigue tests of high-strength steel. Trans ASCE 1961;126:750–63.
713–31. [86] Bannister AC, Skalidakis M, Pariser A, Langenberg P, Gutierrez-Solana F, Sánchez
[51] Filippini M. Stress gradient calculations at notches. Int J Fatigue 2000;22:397–409. L, et al. Performance criteria for cold formed structural steels. European
[52] ISO 965-1:2013. General purpose metric screw threads - Tolerances - Part 1: Commission; 2006.
Principles and basic data. ISO; 2013. [87] Klöppel K. Dauerfestigkeitsversuche mit Schweissverbindungen aus St 52, Stahlbau
[53] Piraprez E. The effect of prying stress ranges on fatigue behaviour of bolted 1960; 1960. p–p.
connections: The state-of-the-art. J Constr Steel Res 1993;27:55–68. [88] Gurney T. Some exploratory fatigue tests on notched mild and high tensile steels.
[54] Schaumann P, Eichstädt R, Oechsner M, Simonsen F. Experimental fatigue Brit Weld J 1964;1964:457–61.
assessment of high-strength bolts with large diameters in consideration of [89] [no names], EUR 5357 e,1975. European Commission; 1975.
boundary layer effects. In: METEC and 2nd ESTAD 2015, Düsseldorf; 2015. [90] Berto F, Mutignani F, Tisalvi M. Notch effect on the fatigue behaviour of a hot-dip
[55] Unglaub J, Reininghaus M, Thiele K. Zur Ermüdungsfestigkeit von feuerverzinkten galvanized structural steel. Strength Mater 2015;47:719–27.
Zugstäben mit Endgewinden. Stahlbau 2015;84:584–8. [91] Berger C, Schaumann P, Stolle C, Marten F. Fatigue strength of high strength bolts
[56] VDI 2230:2015. Systematic calculation of highly stresses bolted joints. VDI; 2015. with large diameters (Ermüdungsfestigkeit hochfester Schrauben großer
[57] Hanenkamp W. Untersuchungen zur Zeit-und Dauerfestigkeit von Hochfesten Abmessungen). Zentrum für Konstruktionswerkstoffe, Technische Universität
Schraubenbolzen (10.9) im Durchmesserbereich M36 bis M72. Konstruktion 1992; Darmstadt; 2008.
44:255–60. [92] Bouwman LP. Fatigue of Bolted Connections and Bolts Loaded in Tension. Stevin
[58] Schaumann P, Eichstädt R. Ermüdung sehr großer HV-Schraubengarnituren. Laboratory, Delft University of Technology; 1979.
Stahlbau 2016;85:604–11. [93] Kuperus A. Fatigue Tests on Tensile Loaded High Strength Bolts and Studbolts.
[59] Van Beek A. Advanced Engineering Design - Lifetime Performance and Reliability. ECCS/Delft University of Technology; 1971.
TU Delft; 2019. [94] Marten F. Zur Ermüdungsfestigkeit hochfester großer Schrauben. Instituts für
[60] James L, Mills W. Review and synthesis of stress intensity factor solutions Stahlbau, Universität Hannover; 2009.
applicable to cracks in bolts. Eng Fract Mech 1988;30:641–54. [95] Schaumann P, Eichstädt R. Fatigue assessment of high-strength bolts with very
[61] Korin I, Perez Ipiña J. Experimental evaluation of fatigue life and fatigue crack large diameters in substructures for offshore wind turbines, in: Proc. In: 25th Int.
growth in a tension bolt-nut threaded connection. Int J Fatigue 2001;33:166–75. Ocean Polar Engng Conf (ISOPE). ISOPE; 2015. p. 260–7.
[62] Leonetti D, Maljaars J, Snijder HH. Fitting fatigue test data with a novel s-n curve [96] Lacher G. Zeit- und Dauerfestigkeit von schwarzen und feuerverzinkten hochfesten
using frequentist and bayesian inference. Int J Fatigue 2017;105:128–43. Schrauben M20 der Festigkeitsklasse 10.9 unter axialer Beanspruchung.
[63] Pascual FG, Meeker WQ. Estimating fatigue curves with the random fatigue limit Bauingenieur 1986;61:227–33.
model. Technometrics 1999;41:277–89. [97] Mang F, Herion S, Fleischer O, Koch E. Untersuchung von Zug-Druck-
Kalottenlagern im Großversuch. Stahlbau 2003;72:43–9.

15