International Journal of Fatigue 72 (2015) 102–108

Contents lists available at ScienceDirect

International Journal of Fatigue

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

Experimental characterization of the bending fatigue strength

of threaded fasteners
Henrik Wentzel ⇑, Xiyue Huang
Department of Solid Mechanics, Royal Institute of Technology (KTH), SE-100 44 Stockholm, Sweden

a r t i c l e i n f o a b s t r a c t

Article history: The fatigue strength in bending of pre-stressed steel bolts is investigated and compared to the fatigue
Received 27 August 2014 strength in axial tension. The strength is measured in terms of maximum engineering stress amplitude,
Received in revised form 31 October 2014 neglecting any stress concentration in the threads. The experimental results reveal that the fatigue limit
Accepted 5 November 2014
is 76% higher in bending than in axial tension. A finite element model is used to compute the stress state
Available online 20 November 2014
in the threaded region for both axial tension and bending. It allows fitting a volume based weakest link
model to the experimentally observed failure probabilities. Based on the good fit of the weakest link
Keywords:
model it is argued that randomly distributed defects in the highly stressed thread root determine the
Fatigue
Bolt
fatigue strength.
Bending Ó 2014 Elsevier Ltd. All rights reserved.
Weakest link
Volume effect

1. Introduction There is no standard test for the bending fatigue of threaded

Bolted joints are widely used for assembly of engineering struc-
tures thanks to their ease of assembly and disassembly and their
low cost. Fatigue failures of bolted joints are often costly and occa-
sionally fatal. This motivates the characterization of the fatigue
strength of bolts.
The testing of the axial fatigue strength of bolt and nut assem-
blies is the subject of international standardization, for example
[1]. Fatigue failures occur primarily in the bolt, the stresses in
the nut are generally much lower. The fatigue strength of the bolt
is expressed in terms of the engineering stress amplitude, e.g. force
amplitude over cross sectional area, thus ignoring the stress con-
centrations in the thread root and in the transition from bolt shank
to bolt head. The fatigue strength in terms of engineering stress can
be compared to loads obtained through handbook calculations or
finite element simulations using the nominal geometry of the bolt,
e.g. ignoring the threads. One such handbook is [2].
In applications many bolts suffer fatigue failure due to bending
loads. This is simply because the joint can support large dynamic
axial forces without generating dynamic stress in the bolt, while
this capacity is much lower in bending [3]. Hobbs et al. performed
fatigue experiments on bolts using eccentric loading [4]. However,
in the eccentric test setup the alternating axial stress is an order of
magnitude larger than the bending stress.
⇑ Corresponding author. Tel.: +46 730469876.
E-mail address: hwentzel@kth.se (H. Wentzel).
fasteners and there is virtually no data on the bending fatigue
strength publically available. This is probably because it is not triv-
ial to construct a testing device that generates a bending load with
high tensile pre-load without using two hydraulic cylinders.
Tests on smooth or notched laboratory specimens typically
show higher fatigue strength in bending than in axial tension [5].
This can be explained by statistical effects, or by considering the
stress gradient [6]. Threaded fasteners have high stress gradients
both in axial tension and in bending due to the geometry of the
thread. It may be meaningful to consider statistical effects based
on the weakest link theory, such as the stressed volume approach
[7] or the stressed area approach [8].
In this paper is presented an experimental setup for bending
testing of threaded fasteners. Experiments in both axial tension
and bending have been carried out on industrial high strength steel
bolts, and the results are reported here. A finite element model of
the bolt and nut assembly is used to resolve the stress state in the
thread root. The stress state is thereafter used to fit a statistical
model based on the stressed volume approach to the experimental
data.
2. Test object
The tested bolts are ISO-metric M14 flange bolts with coarse
thread, strength class 10.9 and length 140 mm, heat treated after
rolling of the threads. All bolts are from the same manufacturer
and the same batch. The bolts are used together with ISO-metric

http://dx.doi.org/10.1016/j.ijfatigue.2014.11.005
0142-1123/Ó 2014 Elsevier Ltd. All rights reserved.
 H. Wentzel, X. Huang / International Journal of Fatigue 72 (2015) 102–108 103

flange nuts of strength class 10. Both nuts and bolt are ordered in
the regular production flow of a vehicle manufacturer. In Fig. 1 is
shown a photographic image of the bolt and nut assembly. Marks
are clearly visible on the ridges of the threads, as seen in the
close-up image in Fig. 2, most certainly due to handling of speci-
men. However, the bottom of the threads appears undamaged.
Three bolts were equipped with strain gauges to enable strain
measurements at the positions 7 mm, 48 mm, and 73 mm from
the bolt head. The strain gauges were positioned on the unthreaded
shank of the bolt and aligned with its axis, one on each side of the
bolt so as to permit measurements of both the axial strain and
the bending strain at the different positions. Thus in total 6 strain
gauges were employed on each instrumented specimen. In the
bending test the bolts were carefully positioned so that the strain
gauges aligned perpendicular to the pivot line so as to measure
the maximum bending strain.
The effective cross-sectional area of the threaded part of the
bolt is Aeff = 115 mm2 [3], corresponding to an effective radius of
reff = 6.05 mm. The effective radius is used to calculate the engi-
neering stress in the threaded part of the bolt. On the non-threaded Fig. 2. Close-up photograph of the threads.
bolt shank the radius is 7.00 mm.

In [9] a distinction is made between the number of cycles to

3. Test equipment and procedures crack nucleation and the number of cycles to failure. No such dis-
tinction is made here, but failure is defined to have occurred when
Testing was conducted at the Laboratory of the Department of the displacement exceeds 5 mm. At this displacement the bolt is
Solid Mechanics in two different servo-hydraulic test machines. broken into two pieces that can be separated by hand.

3.2. Bending testing

3.1. Axial testing

There is no ISO-standard for bending testing of bolts and a novel

Procedures and equipment for axial fatigue testing of bolts are
test setup has been developed in this work. The setup consists of
well described in the standard [1]. The bolts were tested in confor-
two rigid parallel plates that are joined with a hinge. A test
mity with the standard.
machine is used to actuate the hinge, thereby changing the angle
The bolts were inserted through cups that were mounted in the
between the plates. The test specimen (bolt and nut) is positioned
test machine. The machine was a servo-hydraulic MTS with a
through drilled holes in both plates and pre-tension is obtained by
capacity of 160 kN with a control unit from Instron, M8520+. The
applying torque on the nut.
load was measured with the machine’s internal load sensor and
When the hinge is actuated the bolt head and nut head are sub-
the displacement was measured with the machine’s internal LVDT.
jected to forced relative displacement (mainly rotation) and in this
To minimize the effect of any misalignment, spherical washers
way bending stresses are induced in the bolt. In this setup the
were used under both the nut- and the bolt head.
hinge is realized with a hardened steel wedge resting in a groove,
After assembly the bolts were given a pre-tension of 73 kN, cor-
allowing the wedge to pivot around the line of contact. The test
responding to approximately 635 MPa. Note that the yield strength
setup is depicted in Fig. 3 where the most important measures
of the bolt material is at least 900 MPa. Thereafter a time-varying
are indicated.
sinusoidal load of amplitude Fa was added until fatigue failure
It should be noted that the test setup induces both bending and
occurred.
axial stress in the bolt. By positioning the centre of the bolt close to
The engineering amplitude stress is
the line of contact the axial stress amplitude is minimized. How-
Fa ever, the test was difficult to control when the bolt was directly
re ¼ : ð1Þ over the line of contact, so a distance of LB = 1.25 mm was main-
Aeff
tained. In this way the machine could work in tension through
During the testing the stress vary between 635  re to 635 + re. the complete load cycle which proved stable. Shallow slots were
The load ratio, often referred to as the R-value, R = max(r)/min(r), machined in the plates to allow for accurate positioning of the bolt
depends on the stress amplitude. and nut.

Fig. 1. An instrumented specimen of bolt and nut with positions of the strain gauges indicated.
104 H. Wentzel, X. Huang / International Journal of Fatigue 72 (2015) 102–108
Fig. 3. Schematic illustration and photographic image of the bending test setup.
The testing was performed in displacement control to avoid any
undesirable drift in displacement. Thus, a sinusoidal displacement
of amplitude d was applied at the ends of the plates, thereby bend-
ing the bolts. For this purpose a servo-hydraulic MTS machine with
a capacity of 30 kN with a Instron M8500 + control unit was used.
Because the bending moment varies along the bolt, the
engineering stress also varies along the bolt. The measured axial
and bending stresses are shown in Fig. 4 as function of the dis-
placement. It can be observed that the stresses vary almost linearly
with the imposed angle and that the bending stress amplitudes are
a factor 7–10 larger than the axial stress amplitude (depending on
the measurement position).
It is possible to compute the bending stress by modeling the
bolt as a beam with varying cross section subjected to both trans-
versal and normal forces as is done in [10]. The comparatively high
normal force gives a significant contribution to the bending stress
when the bolt deforms. Standard beam theory provides the
governing equation for the transversal displacement w,
w00 ðxÞEI  wðxÞP ¼ M o  Tx: ð2Þ
In Eq. (2) E is the elastic modulus, I is the second moment of inertia
that has a different value in the threaded part than in the bolt
shank, x is the distance along the bolt measured from the bolt head,
Fig. 4. Axial and bending stress in the test setup for different displacements.
P is the pre-tension force, T is the transversal force and Mo is the
bending moment acting on the bolt head. Mo and T are not explicitly
know but can be found by solving Eq. (2) with the boundary
conditions w(0) = w(L) = 0, w0 (0) = h/2, w0 (L) = h/2, where h is the
imposed angle between the bolt- and nut head. This angle is given
by
d
h¼q ; ð3Þ
LA þ LB
where q is a correction factor to account for the compliance of the
test setup. In this work q = 0.85 was used because that value gave
good agreement with the measured strain.
The engineering bending stress amplitude in the bolt is
re ðxÞ ¼ Ew00 ðxÞrðxÞ: ð4Þ
On the bolt shank the calculated engineering stress can be com-
pared to the measurements. One such comparison is shown in
Fig. 5, where the agreement between measured- and calculated
stress in the three measured points is within 3%. It is the stress
in the threaded part that is important because this is where failures
occur most often. However, all the measurements are performed
on the bolt shank, because that is where it is possible to fix sensors.
The sensors are used to validate the computational model for the
engineering stress. When reporting test results the maximum engi-
neering stress amplitude in the threaded region is reported, refer to
Fig. 5. The stresses in the engaged threads are not reported.
The nuts are tightened with torque control to 181 Nm, this
means that the exact pre-tension is not known. For the friction
coefficient of 0.1 in the threads and 0.13 between the nut and
the test plate the expected pre-tension is 73 kN [3]. Several
tightening experiments were performed on the three bolts that
were instrumented with sensors and pre-tensions in the range
72–82 kN were recorded. The difference is primarily due to scatter
in friction. It was noted that after 3–5 tightening of the same bolt
and nut the friction increased significantly, resulting in lower pre-
tension values. For the specimens that were not instrumented the
pre-tension was considered to be 73 kN.
Because the displacement was controlled in the test the stresses
decreased when fracture occurred, thereby reducing the rate of
crack growth. Failure was defined to occur in the load cycle when
the force required for maintaining the constant displacement
amplitude d only reached 90% of the corresponding force in the
beginning of the test.
Fig. 5. Bending stress amplitude in the bolt at displacement d = 1 mm. The threaded
region starts 80 mm from the bolt head as indicated by the overlaid schematic
image of the test specimen.
 H. Wentzel, X. Huang / International Journal of Fatigue 72 (2015) 102–108
3.3. Generation of stress-life diagram
The results of the fatigue testing are summarized in the form of
a Wöhler-diagram that describe the number of load cycles that
specimens survive at a constant stress amplitude, e.g. the fatigue
strength. A S/N-line representing the fatigue strength at 50% failure
probability can be calculated and used for dimensioning bolts in
engineering structures.
For this purpose it is assumed that the S/N line has the form
log10 ðNÞ ¼ alog10 ðSÞ þ b:
In Eq. (5) the parameters a and b are properties of the bolts,
they are fitted (least square fitting) to the test results.
3.4. Estimating the fatigue limit
The fatigue strength at N = 106 is referred to as the fatigue limit,
Su. A commonly used method for determining the fatigue limit at
50% failure probability is the two-point method. This method
makes use of the test data at two different stress levels to
determine the fatigue limit. The two point method is based on
the assumption that the fatigue limit of a population is normally
distributed with mean value l and standard deviation m. The prob-
ability of failure prior to 106 load cycles at load level re is then also
normally distributed
  
1 re  l
Pf ðre ; l; tÞ ¼ 1 þ erf p :
2 t 2
Here, several stress levels are tested and a maximum likelihood
estimation of the fatigue limit, Su, is computed. The interest in
using the maximum likelihood estimate instead of the two-point
method lies in the possibility to consider several test levels,
thereby increasing the confidence in the estimate. To this extend
the likelihood function, L, is defined
(
Y
M
Pf ðrk ; l; tÞ; if specimen no: k failed
e
Lðl; tÞ ¼ : ð7Þ
k¼1
1  Pf ðrke ; l; tÞ; else
In Eq. (7) rke is the engineering stress amplitude of test number k
and M is the total number of tests used for making the estimation.
The likelihood function thus describes the probability of obtaining
the experimental outcome. By maximizing the likelihood function,
estimates of l and v are obtained and used to compute a 95%
confidence interval for the fatigue limit, Su.
3.5. Weakest link model
To investigate if the volume effect can explain the difference
between the fatigue limit in tension and the fatigue limit in
bending a weakest link model has been developed.
The material inhomogeneities are modeled as uniformly distrib-
uted in the bolt with a constant defect density per unit volume. The
defects may for example be non-ferrous inclusions. The stress
required for crack growth is determined by the size of the defect
in the material. Weakest link theory is based on the following
assumptions [11]:
 Cracks originate in defects.
 Fracture of a sub-volume leads to failure of the structure.
 The size of the defects is strictly smaller than the distance
between defects so that no interaction between defects
occurs.
The computations follow those of Weibull’s memorable paper
[12]. The failure probability of the specimen is computed
105
 Z  m 
1 r1  rth
Pf ¼ 1  exp dV : ð8Þ
V ref V rref
where rth, rref and m are constants characteristic of the material.
Vref is a reference volume that has been added to Weibull’s original
works to obtain appropriate dimensions of the ingoing parameters.
Here Vref = 105 m3, which is approximately equal to the volume of
test specimen.
The stress is computed in the entire bolt by means of a finite
ð5Þ element model and the first principal stress, r1, is used in the
Weibull integral Eq. (8). Then the material constants, rth, rref and
m, are fitted to the experimental failure probabilities using a least
square fitting. The computation is done for the combination of
axial- and bending test, weighted with the number of tests at each
load level.
3.6. Finite element modeling
A finite element model was used to compute the stress state in
the specimens. The bolt and nut are modeled in ABAQUS 6.12 with
7  105 linear brick elements, C3D8I. In the highly stressed region
the threads are modeled with an element size of 90 lm corre-
sponding to 10 elements in the thread root radius. The geometry
of the bolt and nut follow the standard [13] with tolerance
6H/6 g as specified in [14,15], the mean values of the tolerance
intervals are used. The bottom surfaces of the bolt- and nut head
ð6Þ
are constrained to two coincident but individual nodes located at
the centre of hinge. These two nodes are used to apply pre-load
and loads on the bolt. In [10] the nut was actually screwed onto
the bolt in the simulation, allowing for an accurate prediction of
clamping force and loosening during subsequent loading. That is
the preferred method but in order to save computational time
the nut is loaded with axial load in this work. The finite element
model is shown in Figs. 6 and 7.
The frictional contact between the threads is modeled
using the standard penalty formulation using a friction coefficient
of 0.1. This value is typical for threads with handling oil and was
also used in Section 3.2 to estimate the pre-tension. An elastic–
plastic material model is employed, incorporating J2-plasticity
with yield stress rs = 900 MPa and linear isotropic hardening. The
run-time of the model is about 80 CPU-hours on a linux-cluster
with 72 cores.
In Fig. 8 is shown the stress distribution in the bolt after pre-
tension. The two most highly stressed areas are the first thread
(after the bolt shank) and the first thread that engaged with the
nut. When bending is applied, Fig. 9, it is also here that the highest
amplitude stresses are obtained. These are the two most probable
locations of failure. It is noted that the most highly stressed regions
suffers plastic deformations already at pre-load. It is thus impor-
tant to employ a plastic material model to accurately compute
the stresses. The plastic strains are small, less than 1%, and the
stress in the thread root has a large hydrostatic (tensile) compo-
nent. The von Mises stress is not above 950 MPa anywhere in the
model.
Fig. 6. Finite element model of the bolt and nut with the bottom surfaces of the
bolt- and nut-head indicated.
106 H. Wentzel, X. Huang / International Journal of Fatigue 72 (2015) 102–108

Table 1
Parameters of the S/N lines of M14/10.9 steel bolts.

Test configuration Axial tension Bending

a 2.81 2.83
b 10.7 11.3
Su 54 ± 3 (MPa) 95 ± 12 (MPa)

failure in the range of 104 to 106 load cycles. In axial testing this
meant stress amplitudes from 41 MPa to 212 MPa, corresponding
to R-values from 0.93 to 0.64, respectively. In bending testing the
stress amplitudes were between 76 MPa and 333 MPa, correspond-
ing to R-values of 0.79 and 0.31, respectively. The test logs are
detailed in Appendix A.

4.1. Stress-life diagram

Fig. 7. Detail of the finite element model.
The principal stresses together with the element volumes are
used for numerical evaluation of the Weibull integral Eq. (8). The
methodology follows the work presented in [7].
4. Results
In total 37 bolts were tested with axial load and 29 bolts were
tested with bending load. Because the mean stress was fixed in
the testing different R-values were obtained for different stress
amplitudes. The stress amplitudes were chosen so as to obtain
The S/N line is parameterized according to Eq. (5). A least square
fit to the test result was performed, and is presented in Table 1. The
fitted S/N lines are displayed in Fig. 10 together with the test data.
It is noted that the slopes of the S/N lines are almost identical while
there is a significant difference in fatigue limit.
No tests were run further than 2 million cycles and this seems
justified by the fact that no failures were obtained between 1
million – and 2 million cycles. In the bending test, 13 out of the
21 failures occurred in the first thread in the transition from shank
to threaded part. The other 8 failures occurred in the first thread
engaged with the nut. In the axial test 26 out of the 27 failures
occurred in the first thread engaged with the nut, the remaining
1 failure occurred in the first thread.

Fig. 8. Distribution of the 1st principal stress in cut-view of the model at pre-tension 73 kN. Limit of the scale is set to 1000 MPa (red). (For interpretation of the references to
colour in this figure legend, the reader is referred to the web version of this article.)

Fig. 9. Distribution of the 1st principal stress in cut-view of the model after pre-tension 73 kN and bending h = 0.01 radians. Limit of the scale is set to 1000 MPa (red). The
radial deformation is exaggerated 20 times. (For interpretation of the references to colour in this figure legend, the reader is referred to the web version of this article.)
 H. Wentzel, X. Huang / International Journal of Fatigue 72 (2015) 102–108 107

Table 3
Material constants for the weakest link model.

Vref (m3) rref (kPa) rth (MPa) m

5
10 49.6 195 1.79

Fig. 10. Wöhler diagram showing the fatigue strength of M14/10.9 bolts, each
marker corresponds to a tested specimen and run-outs are marked with circles.

4.2. The weakest link parameters and predictions

The tests that were performed in the vicinity of the fatigue limit
are detailed in Table 2 and used to fit the statistical model. Based
on the data in Table 2 and the stress computed with the finite ele-
ment model the material constants rth, rref and m are computed.
The constants are presented in Table 3 and visualized in Fig. 11. Fig. 11. Relationship between maximum engineering stress and failure probability
prior to 2  106 load cycles. Solid lines show values computed with the Weibull
It is noted that the failure probability computed with Eq. (8) fits
integral, markers show the test data.
very well with the experimentally observed values for both the
axial tension test and the bending test.

The fatigue limit in bending is usually higher than in axial ten-

5. Discussion
An alternative method for evaluating pre-stressed bolts sub-
jected to alternating pure bending is to use a test machine with
two independently controlled hydraulic cylinders. However, it
has been shown that the proposed single cylinder test generates
bending stresses much larger than the alternating axial stresses
while the mean stress state is dominated by axial stresses. The
bending stress amplitudes are a factor 10 larger than the axial
stress amplitudes in the critical points. Thus the proposed test
setup allows for a fair evaluation of the bending fatigue of pre-
stressed bolts.
It is well known that there can be significant scatter in strength
between different batches of bolts and it is therefore necessary to
perform reference testing in axial tension. For the tested batch of
bolts the fatigue limit in axial tension was 54 ± 3 MPa. According
to [2] a value of 49 MPa is typical for bolts of this type. This differ-
ence of about 10% is within the expected scatter between batches
and suppliers. A fatigue limit close to the typical value indicates
that the test specimens were of a quality representative of indus-
trial standard.
In the experiments the mean stress was fixed at 635 MPa,
because this is deemed most relevant for applications. An unde-
sired effect of a fixed mean stress is that the R-values vary with
the load amplitude.
sion, see for example [5], and it is the main objective of this work to
quantify this difference for pre-stressed high strength bolts. The
fatigue limit in bending was found to be 95 ± 12 MPa, this is 76%
higher than in axial tension. Note that it is the engineering stress
amplitude that is used for reporting the fatigue limit and the S/N
line. It is computed from a model that neglects any stress concen-
tration in the threads and is therefore much inferior to the true
stress amplitude in the thread root.
The majority of the failures in the bending test occurred in the
first thread. This result is unexpected because the engineering
stress amplitude is higher in the first thread engaged with the
nut, refer to Fig. 5. However, the difference in stress is only 3%
and possibly the fatigue failures can be attributed to scatter of
the material properties.
The slope of the bolts’ S/N line was computed to 1/2.8 in both
axial tension and bending. For sharply notched geometries such as
these, values around 1/3 are normal. The fatigue strength in axial
tension at N = 2  105 is 85 MPa (refer to Fig. 10). This value can be
compared to 105 MPa that was reported for a similar bolt (M12/
8.8) in [4]. The difference of about 20% is most probably due to
the scatter between different batches of bolts from different
suppliers.
Weakest link theory has been used to model the experimental
data. The very good fit of the weakest link model, as illustrated
in Fig. 11, shows that the volume effect can fully explain the

Table 2
Test results in the vicinity of the fatigue limit.

Type of test Axial tension Bending

Stress amplitude (MPa) 40.7 47.6 52.0 56.3 65.2 76.0 85.5 95.0 105 109
No. of failures 0 0 1 3 3 0 1 2 2 2
No. of run-outs 4 3 2 1 0 3 2 2 1 0
108 H. Wentzel, X. Huang / International Journal of Fatigue 72 (2015) 102–108

difference in fatigue strength between bending and axial tension. It Table A2

is noted that the threshold stress rth is estimated to rth = 195 MPa, Test results from bending fatigue of M14/10.9 bolts with pre-tension 73 kN
(635 MPa).
suggesting that true stress below this value does not contribute to
the probability of fatigue failure in the bolt. This value is reason- Nr Load amplitude (mm) Engineering stress R-value Cycles to
able, keeping in mind that the yield strength of the material is amplitude (MPa) failure

about 900 MPa. The good fit of the weakest link model indicates B1 2 190 0.54 125,000
a high defect sensitivity of the failure mode, e.g. it is the defects B2 3.5 333 0.31 14,000
B3 1 95 0.74 Run-out
in the thread root that are determining the fatigue strength in B4 1.5 143 0.63 153,000
threaded fasteners. B5 3 285 0.38 24,500
B6 1 95 0.74 Run-out
6. Conclusion B7 1.25 119 0.68 200,000
B8 2.5 238 0.45 29,700
B9 1.1 105 0.72 343,000
The proposed test setup generates bending stress amplitudes in B10 1.25 119 0.68 157,000
the test specimen that are an order of magnitude greater than the B11 1.25 119 0.68 203,600
axial stress amplitudes and can be used for bending fatigue testing. B12 1 95 0.74 469,700
For this batch of M14/10.9 bolts the fatigue limit in bending is B13 1.25 119 0.68 149,300
B14 1 95 0.74 590,000
76% higher than in axial tension.
B15 2 190 0.54 55,000
The higher fatigue limit in bending than in axial tension can be B16 2 190 0.54 53,400
fully explained by Weibull’s weakest link theory, indicating that B17 2 190 0.54 59,570
the defects in the threads are critical for the fatigue strength of B18 0.9 86 0.76 609,000
bolts. B19 0.8 76 0.79 Run-out
B20 0.8 76 0.79 Run-out
B21 0.8 76 0.79 Run-out
Acknowledgements B22 0.9 86 0.76 Run-out
B23 0.9 86 0.76 Run-out
The authors wish to thank Scania AB for financing the study and B24 1.1 105 0.72 480,500
B25 1.1 105 0.72 Run-out
giving the possibility to present these results.
B26 2.5 238 0.45 33,600
B27 2.5 238 0.45 33,100
Appendix A. Test logs B28 1.15 109 0.71 358,000
B29 1.15 109 0.71 510,000

See Tables A1 and A2.

Table A1
Test results from axial fatigue of M14/10.9 bolts with pre-tension 73 kN (635 MPa).

Nr Load amplitude Engineering stress R- Cycles to

Appendix B. Supplementary material
(kN) amplitude (MPa) value failure
A1 8.4 73 0.87 269,347
Supplementary data associated with this article can be found, in
A2 12.6 109 0.81 48,483 the online version, at http://dx.doi.org/10.1016/j.ijfatigue.2014.
A3 10.5 91 0.84 77,162 11.005.
A4 12.6 109 0.81 77,857
A5 12.6 109 0.81 105,082
References
A6 12.6 109 0.81 94,689
A7 4.7 41 0.93 Run-out
[1] ISO 3800:1993. Threaded fasteners – axial load fatigue testing – test methods
A8 6.5 56 0.90 Run-out
and evaluation of results.
A9 6.5 56 0.90 310,761
[2] VDI-2230 Systematic calculation of high duty bolted joints – joints with one
A10 6.5 56 0.90 450,047 cylindrical bolt. Verein Deutscher Ingenieure; 2003.
A11 6.5 56 0.90 808,835 [3] Handbok om skruvförband 2nd ed., Kista: Colly Company AB; 1995.
A12 4.7 41 0.93 Run-out [4] Hobbs JW, Burgete RL, Heyes PF, Patterson EA. The effect of eccentric loading
A13 4.7 41 0.93 Run-out on the fatigue performance of high-tensile bolts. Int J Fatigue 2000;22:531–8.
A14 4.7 41 0.93 Run-out [5] Sundstöm B, editor. Handbook of Solid Mechanics. Stockholm: Department of
A15 14.2 123 0.79 46,907 Solid Mechanics, KTH; 2010.
A16 14.2 123 0.79 63,366 [6] Norberg S, Olsson M. The effect of loaded volume and stress gradient on the
A17 14.2 123 0.79 64,090 fatigue limit. Int J Fatigue 2007;29(12):2259–72.
A18 16.8 146 0.75 41,463 [7] Sadek S, Sandberg D, Olsson M. FE-mesh effect of the volume based weakest-
A19 16.8 146 0.75 40,054 link fatigue probability applied to a compressor blade. In: Proceedings of the
A20 8.4 73 0.87 197,174 ASME Turbo Expo. vol. 7, Issue PARTS A AND B; 2012. p. 427–38.
[8] Sadek S, Olsson M. New models for prediction of high cycle fatigue failure
A21 8.4 73 0.87 477,380
based on highly loaded regions. Int J Fatigue 2014;66:101–10.
A22 5.5 48 0.92 Run-out
[9] Korin I, Perez Ipiña J. Experimental evaluation of fatigue life and fatigue
A23 5.5 48 0.92 Run-out crack growth in a tension bolt-nut threaded connection. Int J Fatigue
A24 5.5 48 0.92 Run-out 2011;33:166–75.
A25 6.0 52 0.91 Run-out [10] Izumi S, Yokoyama T, Iwasaki A, Sakai S. Three-dimensional finite element
A26 6.0 52 0.91 778,492 analysis of tightening and loosening mechanism of threaded fastener. Eng Fail
A27 6.0 52 0.91 Run-out Anal 2005;12:604–15.
A28 7.5 65 0.89 796,679 [11] Nilsson F, editor. Fracture Mechanics – from Theory to Applications.
A29 10.5 91 0.84 140,632 Department of Solid Mechanics, KTH; 2001.
A30 10.5 91 0.84 166,513 [12] Weibull W. A statistical theory of the strength of materials. Ingen-
A31 18.5 160 0.73 31,027 jörsvetenskaps Akademins Handlingar 1939;151.
A32 23.0 199 0.66 20,375 [13] ISO 724:1993. ISO general-purpose metric screw threads – basic dimensions.
A33 7.5 65 0.89 341,317 [14] ISO 965–1:2013. ISO general purpose metric screw threads – tolerances – Part
1: principles and basic data.
A34 7.5 65 0.89 513,857
[15] ISO 965–2:1998. ISO general purpose metric screw threads – tolerances – Part
A35 10.5 91 0.84 150,236
2: limits of sizes for general purpose external and internal screw threads –
A36 18.5 160 0.73 30,565
medium quality.
A37 24.5 212 0.64 15,270