Journal of Constructional Steel Research 167 (2020) 105955

Contents lists available at ScienceDirect

Journal of Constructional Steel Research

Rivet clamping force of as-built hot-riveted connections in steel bridges

Davide Leonetti a,⁎, Johan Maljaars a,b, Giacomo Pasquarelli c, Giuseppe Brando c
a
Department of the Built Environment, Eindhoven University of Technology, the Netherlands
b
TNO, the Netherlands
c
Department of Engineering and Geology, University G. d'Annunzio of Chieti-Pescara, Italy

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

Article history: Hot-riveted connections have been widely used in the past for metallic bridges, of which a large part is still in ser-
Received 23 November 2019 vice, making relevant the assessment of their fatigue life. Previous studies have shown that the fatigue behavior of
Received in revised form 23 January 2020 hot-riveted connections depends on many factors; among these, the residual tensile force in the rivets that
Accepted 27 January 2020
clamps the plates together, i.e. the clamping force, is one of the most prone to uncertainty and scatter. Investiga-
Available online xxxx
tions p in the past made use of specimens produced in controlled laboratory conditions, potentially leading to op-
Keywords:
timistic results. This paper presents an experimental investigation on the clamping force of as-built hot-driven
Hot riveting process rivets extracted from an old steel bridge. On average, the clamping stress was found to be ∼100 MPa and ∼60
Steel rivets MPa, but with large scatter, for two or three plates being clamped, respectively, and for grip length over diameter
Iron and steel bridges ratios close to unity. This significant dependency on the number of clamped plates, as well as the lower values
Clamping force observed as compared to earlier studies, are attributed to larger imperfections in rivets installed in-service, as
Clamping stress compared to the controlled laboratory environment. In addition, a finite element model is presented that simu-
lates the development of the clamping force following the installation of the rivet. The finite element model was
validated on the basis of the experimental data and it appears able to predict the effect of the grip length on the
clamping force. The larger the grip length over diameter ratio, the larger is the clamping force.
© 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://
creativecommons.org/licenses/by-nc-nd/4.0/).

1. Introduction of a cherry red color. After heating, the rivet is inserted into the pre-
pared hole in the plates.
The hot riveting process has been widely used up to about 1970 for 3. The shop head is blocked using a bucking bar with a concave hemi-
wrought iron and steel structures, giving rise to thousands of infrastruc- spherical end, and a second head, the field head, is forged through
tures that currently form a valuable cultural heritage. Many of these are hammering. In the past, the hammering process was carried out by
still in service for railways and roads [1]. Rivets were mostly used for a man holding hammer, later replaced by a pneumatic or a hydraulic
joining plates in built-up cross-sections and for creating structural de- hammer. The hammering also affects the shaft of the rivet, which fills
tails, such as girder to cross beam connections or splice connections. the clearance previously existing with the hole. The original length of
A Rivet is a permanent mechanical fastener, originally consisting of a the shank must, therefore, include the useful length to cover the sum
hemispherical head, called shop head, and a smooth cylindrical shank. of the thicknesses of the parts to be connected, and an extra length,
The hot-riveting process is divided into several phases: commonly called tail [2]. It not only accounts for the volume neces-
sary to forge the field head, but also for the volume of the hole not
1. Holes are created in the plates to be joined. The holes have a properly
filled by the original rivet shank, see Fig. 1. As stated in [3] the rivet
calibrated diameter, greater of at least 1 mm than the diameter of the
reaches a temperature of about 700 °C just after the forging of the
original rivet shank, and are the results of punching processes, some-
second head. At this stage, because of the Young modulus and yield
times followed by rimming, or, in case of plates that are too thick for
stress that the steel exhibits, which are one order of magnitude
punching, the holes are drilled.
lower than at the service temperature, no significant clamping
2. The rivet is heated up to about 900 °C in a special forge. In the past,
force can be established.
since no sufficient measurement instruments were available, this
4. Following the forging phase, cooling of the rivet (in air) takes place:
temperature was assumed to be reached when the rivet appeared
during this phase, both the yield stress and the ultimate tensile
strength of the steel increase and the rivet shank shortens. The com-
bination of these aspects causes that the plates to be joined are
⁎ Corresponding author. pressed against each other by the rivet heads and that an internal
E-mail address: d.leonetti@tue.nl (D. Leonetti). axial force arises in the shaft, known as “clamping force”. Moreover,

https://doi.org/10.1016/j.jcsr.2020.105955
0143-974X/© 2020 The Authors. Published by Elsevier Ltd. This is an open access article under the CC BY-NC-ND license (http://creativecommons.org/licenses/by-nc-nd/4.0/).
2 D. Leonetti et al. / Journal of Constructional Steel Research 167 (2020) 105955

connections [14,18,19,23]. For example, in [23] the fatigue life of riveted

connections was compared with that of bolted connections having nor-
Shop mal and reduced bolt pre-loads. It resulted that bolted connections with
head R
pre-stressed bolts have a higher and less scattered fatigue life than
riveted connections. This was attributed to the higher and less scattered
clamping force that bolts exhibit with respect to rivets.
Few investigations have been performed to evaluate the clamping
force. In early works, it was postulated works that it should reach the
Grip length yield stress of the material [26]. However, as recognized later and sup-
shank
ported by experimental evidence, the clamping force almost reaches
the yield stress only for long grip length (N100mm). Instead, for shorter
grip length the mean of the clamping force decreases, and its scatter in-
tail creases, [4,26,27]. This is attributed to the deformation of the rivet heads,
which has a stronger effect on the clamping force for short grip lengths.
Also, based on the results reported in [4], which were obtained from
rivets having diameter of 24.5 mm (1 in.), the clamping force appears to
depend on the steel type. For this reason, in [24] the ratio between the
nominal tensile stress in the rivet shank cross-section due to the
clamping force, namely the clamping stress, and the measured yield
Fig. 1. Rivet inserted in plate assembly prior to the forging of the field head (displayed in stress of the rivet has been evaluated as one of the meaningful parame-
symmetry). ters affecting the rivet performance. The conducted analyses reported
average ratios of 0.61 and 0.77 for grip lengths of 75 mm and 125 mm,
respectively. In [28] it was amended that the clamping force in speci-
the thermal and lateral contraction in the radial direction of the rivet mens made of a steel type St52 is lower than that one of specimens
shaft creates a gap between the rivet shaft itself and the hole, as ex- made of a steel type St37.
perimentally observed in [4]. All of the aforementioned studies dealt with the clamping force of
rivets installed in a laboratory environment, on relatively long grip
Riveted joints are commonly intended to transfer both shear and length, but, on the other hand, the clamping force in rivets installed in
tensile forces, as for bolted joints. The main difference consists in the cir- bridges, especially those being installed in-situ, may differ from these
cumstance that a rivet is a permanent fastener and cannot be re- experimental values because of imperfections and tolerances in the riv-
tightened afterwards. A relevant issue concerns the assessment of the eting process. In [3], 9 values of the clamping stress are reported. The
remaining working life of these joints in bridges, taking account of fa- riveted specimens were characterized by a presumable grip length of
tigue deterioration under traffic loads [5,6]. Fatigue resistance (S\\N) 30 mm and were taken out from a built-up riveted girder consisting of
curves are traditionally based on regression of fatigue tests, e.g. [3,7], a web plate and four flange angles, which was part of a demolished
and are given in several recommendations -such as JRC-ECCS [8] and bridge. The measured clamping stress was on average equal to
the German RIL805 [9]. On the other hand, recent studies have shown 84.1 MPa with a standard deviation equal to 41.1 MPa. The clamping
to be truly capable in predicting the fatigue performance of hot riveted stress has been evaluated by placing a special strain gauge into a hole
connections using a numerical approach that combines the results ob- drilled into the shaft of the rivet. Then, the rivet was removed from
tained through a finite element model (FEM), implemented in order the plates and the clamping strain in the rivet shank was obtained
to determine the fatigue notch factor, and a strain-life relationship, from the reading of the strain gauges. Other experimental analyses
which is used to relate that fatigue notch factor to the fatigue life were reported by Akesson [24], and Baron and Larson [25] which re-
[10–13], also by means of probabilistic approaches [14–17]. These pre- ported an average trend in line with previous investigations. The rivets
diction models have shown that fatigue life is largely depending on in [24] were extracted from a stringer beam that was in service, whereas
the specific type of riveted connection and its geometry [13]. Hence, no similar information is available for the tests reported in [25]. All ex-
the prediction models are often used for the assessment of individual perimental data show the importance of the grip length on the clamping
bridges with peculiar details. force: a larger grip length increases the average clamping force and re-
In these models, the clamping force exhibited by the rivet is one of duces the scatter, see Fig. 2. Based on different experimental data,
the important variables required for an accurate prediction of the bear- from an unknown source, the same conclusions were drawn in [26],
ing condition of the rivet and, therefore, of the notch effect, i.e. the fa- see Fig. 2. The steel type and hammering process are apparently of
tigue notch factor. A higher clamping force induces a beneficial effect less relevance for the clamping force.
on the fatigue life of shear riveted connections [18–20], as it allows for In [29] a numerical simulation of the hot riveting process was re-
a larger fraction of the applied force to be transferred through friction, ported, using a coupled thermo-mechanical analysis. The thermal re-
thereby reducing the stress concentration at the notch by contributing sults showed a good agreement with the measured thermal behavior
to a lower bearing force of the rivet. This is confirmed by fatigue predic- of the connection. The grip length was 30 mm and the results showed
tion models developed by several authors, see for example [13,14]. The average tensile stresses in the rivet shaft of about 90 MPa, which is in ac-
clamping force not only influences the average fatigue life of riveted cordance with experimental evidence, see Fig. 2. However, the authors
connections but also its variability. In fact, as the coefficient of variation validated only the temperature results of the finite element model,
of the clamping force is relatively large (CoV≥0.3), and this has been and not the clamping force. In [30] a similar FE model was developed.
demonstrated as related to the variability of the technological aspects Also in this case, the clamping force was not extrapolated as a result,
of the hot riveting process [21]. The fatigue resistance of tensile riveted making it difficult for the validation of the structural analysis. In both
connections also benefits from a higher clamping force, because of in- cases, the analyses were carried out for a fixed geometry, therefore
creased stiffness of the plate assembly, thereby reducing the load fluctu- the effect of the input variables of the model was not quantified.
ation experienced by the rivets, as it happens for tensile bolted In this framing of research, the current paper shows the results of an
connections [22]. On the other hand, the high variability in the clamping extensive experimental and numerical investigation carried out with
force is shown to be the main reason for the relatively high scatter in the the aim of quantifying the clamping force in as-built rivet connections
fatigue life of riveted connections when compared to similar bolted of existing steel bridges. Different from most of the previous studies,
 D. Leonetti et al. / Journal of Constructional Steel Research 167 (2020) 105955 3

300 Moreover, the obtained numerical results are used to corroborate and
explain the experimental analyses and to further study the effect of
Clamping Stress [MPa]
250
the grip length on the clamping force.
200
2. Models and methods
150

100 In this section, the methodology implemented for the experimental

tests carried out on 86 specimens extracted by structural details of the
50 Botlek Bridge in Rotterdam is described, followed by the description of
the finite element model used to simulate the development of the
0
20 40 60 80 100 120 140 clamping force that arises after the forging of the field head.
Grip length [mm]
2.1. Measurement procedure for the clamping force
Akesson
Baron & Larson
Wilson and Thomas As already mentioned, the clamping stress in hot-driven steel rivets
Zhou reported in [3] was measured using a special strain gauge inserted in a
Van Maarschalkerwaart hole drilled into the shaft of the rivet. However, no details are available
about the procedure implemented. The current research also used strain
Fig. 2. Experimental results related to the variability of the clamping stress as a function of gauges, and the procedure described in the following has been applied
the grip length for carbon steels. Comparison between experimental data from Akesson
for testing purpose:
[24],Wilson and Thomas [4], and Zhou [3]. Also the average trend from Baron & Larson
[25], and upper and lower bounds from van Maarschalkerwaart [26] are reported.
1. extraction of the specimens from the structural details;
2. instrumentation by installing a special strain gauge in a hole drilled
the clamping force has been measured on rivets from a bridge that was in the center of each rivet shank, as described in [3];
in service for more than 50 years, the Botlek bridge. Given the complex- 3. ejection of the rivet from the specimen and evaluation of the change
ity and the dimensions of the considered structural details from which in the measured axial strain ε;
the specimens have been extracted, the proposed experimental investi- 4. evaluation of the relation between the axial force and the strain mea-
gation is likely to be the first in literature that is addressed to rivets that sured by the strain gauge through an experimental compression test
have been installed in situ. Therefore, these rivets may contain more im- on the ejected rivet;
perfections than those used in previous investigations and conse- 5. correction of the measured strain obtained in step (2), evaluation of
quently, different values of the clamping force than those produced in the residual axial force, and calculation of the residual axial stress.
an ad-hoc controlled laboratory environment. In addition, there is the The structural details out of which the specimens were extracted
possibility that some deterioration occurred due to service conditions. consisted of a K-node of the truss and the wind bracing, which was

Fig. 3. Structural details extracted from the Botlek Bridge in Rotterdam from which the coupons have been obtained.
4 D. Leonetti et al. / Journal of Constructional Steel Research 167 (2020) 105955

located underneath the deck of the bridge, as it is shown in Fig. 3. These

details were extracted using flame cutting, therefore the rivets close to
the cutting paths were not selected for the experimental analyses pro-
posed in this paper because the heat-induced by the cutting procedure
might have affected the residual force, potentially inducing relaxation of
the residual stress state.
Each specimen extracted consisted of a rivet and a portion of the sur-
rounding plate. In this case, the cut was performed using a water jet, in
order to induce as little heat as possible and to do not influence the ele-
ments forming the connection. The cut followed a circular path around
the rivet head. In particular, assuming that the influence of the clamping
stress develops with a conventional angle of 45°, starting from the ex-
ternal edge of the rivet head, see Fig. 4, the minimum diameter that is
necessary for the specimen, in order to minimize the influence of the
cut on the clamping force, is:

d ≥ dmin ¼ g þ a ð1Þ

where g is the grip length and a is the diameter of the rivet head. For
each specimen, the diameter was selected given the value of a and the
grip length g, in accordance with Eq. 1.
As shown in Fig. 5a, in order to carry out the second step of the pro-
cedure, a hole having a diameter of 2 mm has been machined into the
shaft. The length l1 has been imposed so to position the gauge center
of the strain gauge in the center of the rivet shaft. Given the specifica-
tions of the strain gauges BTM-6C type used for the current analysis, a
value of 8 mm has been considered for l1. As it is shown in Fig. 5, a
shorter hole contained in the rivet head and having a diameter equal
to 5 mm has been drilled to facilitate the operations. Moreover, a longi-
tudinal slot has been produced in order to accommodate the gauge
leads. In order to install the strain gauge, the hole was first thoroughly
cleaned with an air jet and a liquid solvent; then, it was filled with a spe- Fig. 5. Drawing of the specimen machined for the installation of the strain gauge (a), and
cial adhesive. Therefore, the strain gauge was inserted, leaving 3 mm picture of the specimen after the installation of the strain gauge.
from the end of the hole, according to the producer specification. Suc-
cessively, the adhesive was first cured for 12 h at room temperature,
and then 4 h at 140 °C. In this phase, rapid variations of temperature measured strain in the rivet at the beginning and after complete ejec-
have been avoided so to exclude air bubbles or cracks in the adhesive tion, is proportional to the relaxation of the rivet and gives an evaluation
provoked by shrinkage at high strain rate. of the level of pre-stress. Due to the fact that the strain gauge is sub-
After installing the strain gauge and connecting it to the control unit, merged in the adhesive, which does not necessarily have the same stiff-
the second head has been removed using a band saw, and water as cool- ness of the steel, the strain value cannot be directly related to the
ant liquid (step 3 of the procedure). Successively, the rivet was nominal longitudinal stress in the rivet shank by using the Hook law.
completely ejected from the plate assembly using a manual jack- Therefore, a calibration procedure was implemented in order to evalu-
hammer. The quantity Δε, i.e. the recorded difference between the ate the relation between the measured strain and the axial force in the
rivet shank.
In the calibration step (step 4 of the procedure), the rivet has been
loaded in compression by a jack-hammer equipped with a load cell,
allowing the determination between the applied axial force, Fcal, and
the measured strain, Δε. In particular, the rivet was equipped with a
hinge in order to avoid secondary bending, see Fig. 6a. The goal was to
obtain a linear relationship between the applied axial force and the
measured strain and to evaluate the compliance as the ratio between
the measured strain and the applied force [μm/(mm kN)]:

Δε
k1 ¼ ð2Þ
ΔF cal

where Fcal is the compression force applied in the calibration phase.

The 5th step of the procedure involves a further correction due to the
circumstance that at the beginning of the test when the clamping force,
Fcl, is established, the rivet is loaded differently than during the calibra-
tion phase when the calibration force, Fcal, is applied These two load
cases are depicted in Fig. 6b: in service the force is transferred by loading
the flat surfaces of the rivet head, whereas during the calibration step
the rivet head is loaded in compression by applying the force on the
head of the rivet. The different stress concentrations in these cases
Fig. 4. Definition of clamping area. due to the different loading conditions may influence the strain at the
 D. Leonetti et al. / Journal of Constructional Steel Research 167 (2020) 105955 5

gauge location since for relatively short rivet lengths the strain gauge is for a rivet shank diameter dr = 22 mm. It resulted that the following
placed in a region of stress concentration. Therefore, it is necessary to functional equation is suitable to describe the trend of k2 as function
evaluate a second correction factor, named k2, which takes into account of the grip length:
the difference in the load type. This is estimated through a finite ele-
ment model of the rivet. This finite element model consists of a rivet k2 ¼ 1 þ 6686g −3:433 for g ≥ 10 mm ð3Þ
modeled using axisymmetric elements, considering the geometrical
symmetry with respect to the plane perpendicular to the axis of the where g should be applied in [mm]. The equation indicates that k2 de-
shaft. As for the mechanical modeling, a linear elastic material model creases with increasing grip length. The comparison between the finite
with Youngs modulus equal to 2E+05 MPa and a Poissons modulus element results and the equation is depicted in Fig. 7. The maximum de-
equal to 0.3 has been considered. The scheme depicting the two loading viation is 5.67E-02. Therefore, the strain range value, Δε, acquired from
cases is in Fig. 6b. The first loading case consists of a compressive load the testing phase has to be divided by k2 in order to obtain the average
applied to the rivet head, which simulates the calibration test, see Fig. strain value, which in the calibration is correlated to the applied load.
6a. In the second loading case, a load has been applied to the portion Following these considerations, the relation between the measured
of the surface of the rivet head, which was constrained by the contact strain range and the clamping stress, σc, is:
with the plates, determining the rivet shank to be in tension, see Fig.
6b. Therefore, the rivet has been constrained only by its symmetric 4Δε
σc ¼ ð4Þ
boundary conditions. The average strain in the axial direction, evaluated k1 k2 πd
2

along a path of nodes positioned in the same location where the gauge
strain gauge is positioned, has been evaluated from the results of the fi-
nite element model. The factor k2 has been considered as the ratio be-
2.2. Numerical model
tween the average strains resulting from the two load cases. Following
a parametric study, executed by varying the grip length between 10
The numerical analysis to simulate the development of a clamping
and 50 mm, k2 [−] has been calculated as function of the grip length g
force after installation, i.e. during cooling of the rivet, involved a fully
coupled thermo-mechanical Finite Element model, implemented in
ABAQUS environment [31]. This is a different and much more complex
model than the one described in the previous section, which was used
solely to determine the linear elastic stress concentration. The analysis
described in the current section is able to consider the interaction be-
tween temperature and displacement fields and vice versa. The finite el-
ement model considers a geometry consisting of two plates clamped by
a rivet having both heads already formed. Thus, the effect of the ham-
mering process has not been considered in the numerical determination
of the clamping force. This assumption has been justified by the fact that
the driving process and the hot forming of the field head are executed
when the temperature in the rivet is above 700°C [3]. At this tempera-
ture, both the yield stress and the Young Modulus of the steel are con-
siderably lower than at room temperature [32], this entailing that the
stresses that correspond to the developing strains can be neglected.
The geometry is modeled considering three values of the grip length,
namely 25, 31.8, and 50 mm. The rivet geometry has been constructed
according to the type A defined in the Dutch standard NEN 667:1944,
see Fig. 8. The diameter of the rivet hole is set equal to the 22.2 mm,
which equals the diameter of the rivet after forging the rivet head,
when the rivet is assumed to completely fill the hole as a consequence
of the hot forging process. The size of the plates, which is five times
the grip length, has been set to minimize the boundaries effects on the
analysis. Because of the computational effort required to perform a

3.5
FE results
3 Equation (3)
Correction factor k 2 [-]

2.5

1.5

0.5
20 40 60 80 100
Grip length [mm]
Fig. 6. Rivet during the calibration step (a), and loading schemes of the rivet while
clamping the plates and during the calibration step (b). Fig. 7. Comparison between the results of the FE model and Eq. 3.
6 D. Leonetti et al. / Journal of Constructional Steel Research 167 (2020) 105955

Fig. 8. Nominal geometry of the rivet according to NEN667:1944. Units in [mm].

coupled thermo-mechanical analysis involving material and geometric

nonlinearities, that are necessary to simulate the cooling phase of the
rivet after forging of the field head, 2D axisymmetric elements have
been used. The mechanical properties of the rivet have been assumed
as the nominal values available for St44 steel grade, since, as indicated
in the Dutch standard NEN 667:1944, this is the steel grade to be used
for rivets. Furthermore, it has been imposed that these mechanical fea-
tures vary with the temperature, according to the rules given by EC3
Part 1–2 [32]. The resulting values of the Young modulus E, yield stress
Fig. 9. Mesh of the finite element model.
σy, specific heat Ca, thermal conductivity λa, and Poisson modulus ν are
reported in Table 1 as a function of the temperature T.
The 8-node axisymmetric thermally coupled quadrilateral, biqua-
dratic displacement, bilinear temperature, reduced integration, 3. Results
CAX8RT element, available in the ABAQUS library has been used to
discretize the geometry. A preliminary sensitivity analysis has been car- The results of the experimental tests and of the numerical simulation
ried out in order to establish the mesh sizes, which have been reduced are described in this section.
in correspondence of the contact and stress concentrators, see Fig. 9.
As far as the rivet-to-plates contact properties are concerned, a tan-
3.1. Experimental results
gential behavior, with a penalty type formula and a friction coefficient
equal to 0.3 [2], has been considered. On the other hand, a Lagrange for-
The measurement equipment involved in the experimental tests de-
mulation has been used to model the normal behavior of each contact
scribed in Section 2.1 allowed to monitor the development of the ac-
pair. This formulation avoids penetration, even if a higher computa-
quired strains during testing, according to the trend depicted
tional effort is required. Finally, the thermal conductance that simulates
qualitatively in Fig. 10. Following an initial increase of the axial strain
the steel-air-steel interaction with zero and non-zero contact gap be-
due to the clamping action of the band saw machine, applied to hold
tween the rivet shaft and the edges of the holes, has been selected ac-
the specimen in place, an initial increase in the strain is measured as
cording to the values given in Table 2, which have been established
the cut advances closer towards the center of the rivet shank (a).
according to [33].
Then, during the further evolution of the cut carried out through the
band saw, the measured strain decreases (b) until it stabilizes after
Table 1 the complete separation of the shank (c). Successively, disturbance in
Variation of the mechanical and thermal material properties with temperature [32].

T E σy Ca λa ν
[°C] [MPa] [MPa] [J/Kg°C] [W/m°C] Table 2
Thermal conductivity of steel to steel contact as a func-
20 210,000 277.3 439.8 53.33 0.2837 tion of the gap between the contacting surfaces, values
100 210,000 277.3 487.6 50.67 0.2868 from [33].
200 189,000 277.3 529.8 47.34 0.2906
300 168,000 277.3 564.8 44.01 0.2943 Gap [mm] λa [W/m°C]
400 147,000 277.3 605.9 40.68 0.2981
0.0 2500
500 126,000 216.3 666.5 37.35 0.3050
0.5 50
600 65,100 130.3 760.2 34.02 0.3142
1.0 25
700 27,300 63.78 1008 30.69 0.3234
1.5 0
 D. Leonetti et al. / Journal of Constructional Steel Research 167 (2020) 105955 7

the signal is recorded during the ejection of the rivet (d). Finally, the 200

Measured strain [ m/mm]

value of the strain stabilizes at a lower level of the measured strain 0
after complete ejection (e). The trend described above was recorded
for most of the rivets, as it is shown in Fig. 11, for the rivets extracted -200
from the wind bracing detail having grip length g = 50 mm. -400
In some cases, the strain measured at the end of the stage (c),
i.e. before the complete ejection of the rivet, equals the strain mea- -600
sured at the end of the stage (e), i.e. after the complete ejection. -800
This indicates that a gap between the rivet shank and the rivet
hole was present at the beginning of the test, and that the gap -1000
0 0.2 0.4 0.6 0.8 1
was larger than the lateral expansion that the rivet shank experi- Normalised time [-]
ences due to the reduction of the internal tensile force. This is the
case for the specimen 3, see Fig. 11. In other cases, the absolute Specimen 1
value of the strain measured at the end of the stage (c), is smaller Specimen 2
Specimen 3
than the absolute value of the strain measured at the end of the Specimen 4
stage (e). This is sufficient to demonstrate that either the gap was Specimen 5
not present, or that the gap was smaller than the lateral expansion Specimen 6
caused by the reduction of the internal tensile force occurring due Specimen 17
Specimen 18
to the removal of the head.
For most of the specimens, some amount of paint came out from the
Fig. 11. Strains measured during the tests for the specimens extracted from the wind
plate to plate contacting surfaces during the hot curing of the strain
bracing, see Fig. 3d, having g = 50 mm. The signals have been zeroed at the start of the
gauge adhesive prior to testing, see Fig. 12a. This indicates that the test and the horizontal axis has been normalised.
red‑lead paint that was present at the contacting surfaces of the joined
plates was still wet despite the decades of service and exposure to
weather conditions, suggesting that the details are air-tight. Fig. 12b de-
Following the procedure described in Section 2.1, the measured
picts the appearance of one of the two plate-to-plate contacting sur-
strains are transformed into stresses, after calibration and adjusted by
faces. It can be noted that the presence of the paint is compromised
applying the correction factor k2, obtained using Eq. 3.
close to the hole, probably due to the fact that the hot rivet was inserted
The obtained test results are presented in Fig. 13 and summarized in
in the hole. As this influences the friction condition between the plates,
Table 3, where each row provides the results of all tests carried out with
it might also influence the effect of the clamping force on the fatigue re-
sistance of shear riveted connections.

Fig. 10. The qualitative trend of the strain acquired by the strain gauge during the test:
during sawing at different cutting depth (a–c), and during (d) and after (e) the Fig. 12. Status of the red lead paint after curing. (For interpretation of the references to
complete ejection of the rivet. color in this figure legend, the reader is referred to the web version of this article.)
8 D. Leonetti et al. / Journal of Constructional Steel Research 167 (2020) 105955

a specific geometry. The entire dataset is reported in the Appendix. Table 3

Also, in Fig. 13 the obtained clamping stresses are compared with Summary of the clamping stress and estimation of the variability. The expected
clamping stress, E[σc], coefficient of variation, CoV[σc], and standard error of the mean
the available results from the literature, significantly corroborating value, St.err.[E[σc]], are reported for each grip length, g, indicating the number of plates
the existing data for short grip length, ranging from 16.8 to and the diameter of the rivet, d.
52.6 mm. With respect to the median trend measured by Baron and
g # of d E[σc] CoV[σc] St.err.[E[σc]]
Larson [25] the present data show lower values. This is possibly
[mm] plates [mm] MPa [−] MPa
due to a difference in the steel grade and in the circumstance that
16.8 2 21.0 102 0.59 15.5
the present specimens are extracted from a bridge that has been in
24.5 2 22.8 120 0.24 7.30
service for decades before being dismissed, whereas Baron and 27.0 2 23 72.1 0.57 16.8
Larson used newly fabricated specimens. On the other hand, the 39.5 2 23.2 120 0.26 15.7
present data are well in agreement with the results of Zhou [3], 50.0 2 22.8 104 0.24 10.1
which were produced with a similar experimental technique and 36.7 3 23 51.8 0.74 8.78
52.6 3 25.7 65.6 0.43 11.5
by using specimens extracted from a Bridge that has been in service
for long times, differently from the results of Wilson and Thomas [4]
that were produced using specimens fabricated ad-hoc.
In agreement with the literature, the test results show an increasing
given for a reduced grip length highlight that a probabilistic description
trend and a decreasing scatter with the grip length. However, from
of the data is not easy and that the probability distribution best describ-
Table 3 it can be recognized that the number of plates affects the aver-
ing the data must necessarily satisfy the asymmetry of the distribution
age value of the clamping force as well as its variability. This is possibly
of the clamping force at relatively low grip length, because σc ≥ 0,
due to imperfections, mainly holes misalignment, which has been found
as well as the asymmetry recorded at relatively high grip length, since
to be frequent in the specimens with 3 plates. Moreover, the results are
σc ≤ σy.
In 14 of the 86 cases, the recorded values of the strain and, therefore,
the results of the test deviated from the results shown in Fig. 10 and Fig.
11. This failed tests are also reported in Table A.6 of the Appendix. In a
few of these cases, either the measured strain increment was positive
suggesting that the installation of the strain gauge was not successful
or the signal was lost during the test. In most of these cases, the strain
increased after ejection, i.e., with reference to Fig. 10, the strain mea-
sured at stage (e) was larger than measured at stage (c). This mostly
happened for very short rivets, where the ejection process might have
damaged either the glue or the glue-steel bond because of the short lig-
ament between the strain gauge hole and the executed cut. These cases
are not shown in Fig. 13. The lower clamping and large scatter encoun-
tered for specimens having three plates have been addressed to the mis-
alignment of the holes of the plates, which determined a deformation of
the rivet shank during the driving process. The shaft of the rivets ap-
peared to have an imperfection on straightness as indicated in Fig. 14.
Such imperfection may be expected in a certain fraction of the rivets
in joints produced in-situ, where the hole has not been correctly
rimmed before the driving process, and it is less likely for joints pro-
duced in shop. In these cases, as the deformation of the rivet shank af-
fected the contraction of the rivet during the cooling phase, a portion

Fig. 13. Clamping stress resulting from the new and the old datasets. Fig. 14. Typical imperfections encountered on the rivet shank.
 D. Leonetti et al. / Journal of Constructional Steel Research 167 (2020) 105955 9

300

250

Clamping Stress [MPa]

200

150

100
Previous works
50 Current work
FE model
0
1 2 3 4 5
Grip length / Rivet diameter

Fig. 15. Comparison of the trend of the temperature measured in [29] with the numerical Fig. 16. Comparison between the average clamping force resulting from the experimental
trend of the temperature obtained by the proposed thermo-mechanical finite element analyses in the present work (only 2 plates), in the previous literature [3,4,24], and the
model. finite element model. In this case, the box-plot diagram indicates for each dataset the
mean value and the 95% confidence interval of the mean.

of the clamping force is likely to be transferred to the plates at the mis-

alignment location. This determines a non-uniform axial force in the In Table 4, the obtained value of the clamping force has been com-
rivet, making the position of the strain gauge along the shaft of the pared with the experimental results of Zhou [3] and Akesson [24]. It re-
rivet relevant but not definable a priori. sults that the current model is able to quantify the clamping force.

3.2. Numerical analyses 3.2.2. Numerical results

The FE model has been used in order to quantify the effect of the
The FE model described in Section 2.2 has been validated consider- grip length on the clamping force. To do so, the grip length has been
ing studies in the literature that faced peculiar aspects, such as the tem- assumed as an input parameter of the model and the resulting
perature field and the strain/stress that arise during the cooling phase. clamping force has been extracted from the results of the thermo-
Furthermore, the model has been used to simulate the development mechanical FE model. Three values of the grip length g have been
of the clamping forces obtained by the experimental tests mentioned used: 25, 31.8, and 50 mm. The resulting values of the clamping
in this paper. In particular, for as far as the thermal results are con- force are plotted in Fig. 16, together with the experimental results,
cerned, the comparison has been carried out monitoring the trend of as a function of the ratio between the grip length and the diameter
the temperature during the cooling phase in the rivet head and on the of the rivet. The results of the model are in good agreement with
plate. Instead, for the validation of the model with respect to the values the trend identified by some of the experimental datasets presented
of clamping forces, the grip length of 30 mm with studied in [3] and the in the paper. In particular, the model is able to predict the mean
different values of grip length studied in this paper have been trend given by the tests which are believed to be carried on speci-
considered. mens having small imperfections. In particular, the model predic-
tion is within the 95% confidence interval of the mean response
for these dataset, which is also summarized numerically in
3.2.1. Model validation
Table 5. The discrepancy with respect to the other datasets is
The trend of the temperature over time in two positions has used
attributed to (1) the effect of the execution of the hammering
for validating the thermal results. The comparison between the re-
process, (2) initial temperature, (3) material properties, and
sults given by the numerical model and the experimental tests is
(4) imperfections.
shown in Fig. 15. Measurements of the temperature profile are
available in [29] for the tip of one of the two rivet heads, and a
point on the plates 30 mm far from the rivet head, see Points A 4. Conclusions
and B in Fig. 15.
As it is possible to observe, the numerical model is able to well repro- Given the results of the present work, the conclusions that follow the
duce the cooling phase observed during the tests, especially with regard performed experimental and numerical analyses are:
to the rivet. With respect to Point B, the used values for the conductance
at the contact, see Table 2, are likely to be the source of bias. After the
cooling phase is complete, the clamping force has been obtained by in- Table 5
tegrating the stress results in the midsection of the rivet shank. The ob- Expected clamping force, E[σc], as a function of the ratio of the grip length over the rivet
tained value of the clamping force for a grip length of 31.8 mm and a diameter, g/d, and 95% confidence interval (C.I.) - 2.5% and 97.5% confidence bounds -
rivet shank diameter equal to 22.1 mm is 127 MPa. for the mean response.

Ref. g/d E[σc] 95% C.I.

[−] [MPa] [MPa]
Table 4 Current Work 0.800 102 [69.4;136]
Expected clamping stress, E[σc], resulting from the finite element model compared with 1.07 120 [105;136]
some experimental results at a close g/d value. 1.17 72.1 [31.1;113]
1.70 120 [76.3;164]
Ref. g d g/d E[σc]
2.19 104 [79.0;128]
[mm] [mm] [−] MPa
[3] 1.36 84.1 [53.1;115]
[24] 30 20 1.5 151 [24] 1.50 151 [117;184]
[3] 30 22.1 1.36 84 [4] 2.86 210 [185;235]
FE model 31.8 22.1 1.43 125 4.76 241 [234;248]
10 D. Leonetti et al. / Journal of Constructional Steel Research 167 (2020) 105955

• A procedure to measure the clamping force in hot-driven steel rivets mised, with a fair chance of affecting the friction coefficient be-
has been described. Different from other studies, a conversion factor tween the plates. This has been addressed to the high
for short grip lengths has been introduced to correct for dissimilar temperatures induced by the hot driving process close to the
stress concentration factors existing between the initial status and rivet shank. Also, the presence of wet paint demonstrates that
the calibration phase, making the proposed procedure more accurate. the joints are almost air-tight, so that corrosion is not expected
• Consistently with previous experimental results, it has been found to be a major problem for fatigue cracks starting from the rivet
that the rivet clamping force increases with the grip length. However, holes, not even in joints without paint.
the effect of imperfections present in real structures limits this effect. • The proposed finite element model allows to well predict the ther-
The clamping stress was approximately 100 MPa on average for a grip mal and mechanical behavior of the rivet after installation. By
length over diameter ratio close to unity and two plates clamped. using the finite element model it is possible to study the effect of
However, the results are found to be dependent on the number of the clamping force as a function of the grip length. The clamping
plates. For three plates clamped, the value drops to approximately force increases with about a factor of two if the grip length over
60 MPa. This has been attributed to the misalignment of the holes in rivet diameter increases from 1 to 2.5. The model results in a
the plates to be clamped, determining a reduction of the clamping good accuracy with some of the test datasets, being bounded be-
force and an increase in its scatter. The coefficient of variation is ob- tween the 95% confidence interval of the mean response.
served to be higher with a factor between 2 and 3 when three plates • The results of this research can be used in the fatigue assessment
are considered instead of 2, as effect of the hole-misalignment. of a riveted connection.
• With respect to data generated using specimens produced in a con-
trolled laboratory environment, the present dataset shows lower Acknowledgements
values of clamping force. This is addressed to a higher incidence of im-
perfections, such as hole-misalignment, head eccentricity, hammering The authors would like to thank the Dutch infrastructural asset
time, and temperature profile, which might negatively affect the owner Rijkswaterstaat for its support.
clamping force and can be determined only if test data are produced
by rivets driven in realistic conditions. Declaration of Competing Interest
• In the case of assemblies in which red lead paint was used to pro-
tect the material from corrosion, such paint has been found to be The authors declare that they have no known competing financial
still wet. However, in the immediate vicinity of the rivet hole, interests or personal relationships that could have appeared to influ-
the integrity of the paint layers has been found to be compro- ence the work reported in this paper.

Appendix A. Experimental results

The appendix summarizes the entire set of results.

Table A.6
Complete set of experimental results.

Specimen Δε ∗ Δε ∗∗ g d # of note k1 note

[μm/m] [μm/m] [mm] [mm] plates [μm/(m kN)]

1 −3.78E+02 −5.09E+02 50 22.8 2 8.26E-02

2 −6.73E+01 −4.98E+02 50 22.8 2 8.47E-02
3 −3.33E+02 −4.09E+02 50 22.8 2 7.52E-02
4 −2.38E+02 −4.50E+02 50 22.8 2 8.55E-02
5 – – 50 22.8 2 (a) –
6 – – 50 22.8 2 (a) –
7 −5.91E+02 −6.55E+02 24.5 22.8 2 1.27E-01
8 −5.37E+02 −6.14E+02 24.5 22.8 2 1.06E-01
9 −4.50E+02 −5.53E+02 24.5 22.8 2 9.62E-02
10 −5.23E+02 −5.88E+02 24.5 22.8 2 1.05E-01
11 −4.45E+02 −4.97E+02 24.5 22.8 2 1.11E-01
12 −3.97E+02 −4.97E+02 24.5 22.8 2 9.43E-02
13 −5.61E+02 −6.07E+02 24.5 22.8 2 9.52E-02
14 −3.81E+02 −4.58E+02 24.5 22.8 2 1.11E-01
15 −4.19E+02 −5.16E+02 24.5 22.8 2 1.12E-01
16 −6.77E+02 −7.18E+02 24.5 22.8 2 1.04E-01
17 −4.09E+02 −7.08E+02 50.0 22.8 2 8.70E-02
18 −3.84E+02 −4.18E+02 50.0 22.8 2 9.90E-02
19 −4.06E+02 −5.07E+02 24.5 22.8 2 9.43E-02
20 −5.89E+02 −6.32E+02 24.5 22.8 2 8.33E-02
21 −1.70E+02 −2.75E+02 24.5 22.8 2 9.80E-02
22 −3.67E+02 −4.46E+02 24.5 22.8 2 9.35E-02
23 −4.14E+02 −5.09E+02 24.5 22.8 2 1.12E-01
24 −2.79E+02 −4.20E+02 24.5 22.8 2 9.80E-02
25 −1.51E+02 −2.82E+02 27.0 23.0 2 9.17E-02
26 −1.98E+02 −1.70E+02 27.0 23.0 2 7.19E-02
27 −3.50E+02 −3.70E+02 27.0 23.0 2 1.05E-01
28 −2.06E+02 −3.47E+02 27.0 23.0 2 1.03E-01
29 −3.53E+02 −5.10E+02 27.0 23.0 2 1.25E-01
30 −2.03E+02 −2.89E+02 27.0 23.0 2 6.17E-02
31 – −4.00E+02 27.0 23.0 2 (c) –
32 −4.53E+01 −1.15E+02 36.7 23.0 3 6.21E-02
 D. Leonetti et al. / Journal of Constructional Steel Research 167 (2020) 105955 11

Table A.6 (continued)

Specimen Δε ∗ Δε ∗∗ g d # of note k1 note
[μm/m] [μm/m] [mm] [mm] plates [μm/(m kN)]

33 −7.82E+01 −1.97E+02 36.7 23.0 3 6.76E-02

34 −7.41E+01 −1.00E+02 36.7 23.0 3 7.41E-02
35 −7.83E+01 −1.41E+02 36.7 23.0 3 6.41E-02
36 −7.75E+01 −1.17E+02 36.7 23.0 3 7.75E-02
37 −4.31E+02 −4.54E+02 36.7 23.0 3 8.20E-02
38 −3.21E+02 −3.33E+02 36.7 23.0 3 8.40E-02
39 −1.62E+02 −2.41E+02 36.7 23.0 3 8.93E-02
40 −1.53E+02 −2.76E+02 36.7 23.0 3 6.17E-02 (d)
41 −5.37E+02 −6.08E+02 36.7 23.0 3 1.11E-01
42 −7.29E+01 −1.88E+02 36.7 23.0 3 9.09E-02
43 −1.68E+02 −2.73E+02 36.7 23.0 3 7.14E-02 (d)
44 −2.43E+02 −3.07E+02 36.7 23.0 3 8.62E-02
45 −1.55E+02 −2.86E+02 36.7 23.0 3 8.06E-02 (d)
46 −1.62E+02 −5.60E+02 36.7 23.0 3 1.00E-01
47 – – 36.7 23.0 3 (c) (a) –
48 −4.34E+01 −4.86E+01 36.7 23.0 3 8.55E-02
49 −8.01E+01 −2.28E+02 36.7 23.0 3 9.43E-02
50 −6.41E+01 −1.55E+02 36.7 23.0 3 8.20E-02
51 −1.24E+02 −2.21E+02 36.7 23.0 3 1.04E-01
52 – – 16.8 21.0 2 (a) –
53 – −4.48E+02 16.8 21.0 2 (b) 1.43E-01
54 −3.83E+02 −3.98E+02 16.8 21.0 2 1.49E-01
55 – – 16.8 21.0 2 (a) –
56 −5.01E+02 −4.43E+02 16.8 21.0 2 (b) 1.85E-01 (d)
57 −4.55E+02 −4.93E+02 16.8 21.0 2 1.16E-01
58 −3.87E+02 – 16.8 21.0 2 (b) – (f)
59 −8.21E+02 −7.96E+02 16.8 21.0 2 9.52E-02
60 −8.63E+02 −9.70E+02 16.8 21.0 2 1.56E-01
61 −3.09E+02 −3.53E+02 16.8 21.0 2 1.64E-01
62 – −9.41E+01 16.8 21.0 2 (c) 9.52E-02
63 −3.81E+02 −3.88E+02 16.8 21.0 2 2.00E-01
64 – – 16.8 21.0 2 (a) –
65 −2.02E+01 – 16.8 21.0 2 (b) 1.33E-01
66 – – 16.8 21.0 2 –
67 −3.74E+02 −4.14E+02 16.8 21.0 2 1.02E-01 (e)
68 −3.81E+02 −3.85E+02 16.8 21.0 2 2.08E-01
69 −2.64E+02 −9.80E+01 16.8 21.0 2 (b) – (g)
70 −2.23E+02 −3.55E+02 16.8 21.0 2 7.09E-02 (e)
71 −7.80E+01 −4.40E+01 16.8 21.0 2 (b) 1.82E-01
72 −6.06E+02 −6.30E+02 16.8 21.0 2 – (g)
73 −3.24E+02 −3.61E+02 16.8 21.0 2 – (g)
74 −1.19E+02 −3.85E+01 16.8 21.0 2 (b) 1.41E-01
75 −1.21E+02 −2.22E+02 16.8 21.0 2 9.52E-02
76 −5.60E+02 −6.68E+02 39.5 23.2 2 8.85E-02
77 −4.31E+02 −6.76E+02 39.5 23.2 2 8.47E-02
78 −2.59E+02 −3.59E+02 39.5 23.2 2 8.77E-02
79 – – 39.5 23.2 2 (a) 1.18E-01 (d)
80 −6.06E+02 −7.24E+02 39.5 23.2 2 8.20E-02
81 −4.38E+01 −2.50E+02 52.6 25.7 3 9.09E-02
82 −1.71E+02 −2.83E+02 52.6 25.7 3 8.62E-02
83 −1.96E+02 −2.70E+02 52.6 25.7 3 1.79E-01
84 −1.50E+02 −3.93E+02 52.6 25.7 3 1.18E-01
85 −8.59E+01 −1.41E+02 52.6 25.7 3 1.16E-01
86 −2.51E+02 −4.37E+02 52.6 25.7 3 1.10E-01

(*) Evaluated before the ejection of the rivet, i.e. stage (c) of the procedure as depicted in Fig. 10
(**) Evaluated after the ejection of the rivet, i.e. stage (e) of the procedure as depicted in Fig. 10
(a) Positive Strain
(b) Increase in strain after ejection
(c) Disturbed signal during test
(d) Evaluated in the range between −1 to −5 kN
(e) Evaluated in the unloading
(f) Damaged
(g) Unreadable

References
[6] G. Stamatopoulos, Fatigue assessment and strengthening measures to upgrade a
[1] Q. Collette, S. Sire, W.J. Vermes, V.J. Mesler, I. Wouters, Experimental investigations steel railway bridge, J. Constr. Steel Res. 80 (2013) 346–354.
on hot-driven structural rivets in historical French and Belgian wrought-iron struc- [7] A. Taras, R. Greiner, Development and application of a fatigue class catalogue for
tures (1880s–1890s), Constr. Build. Mater. 54 (2014) 258–269. riveted bridge components, Struct. Eng. Int. 20 (1) (2010) 91–103.
[2] J.W. Fisher, J.H. Struik, Guide to Design Criteria for Bolted and Riveted Joints, 1974. [8] B. Kuehn, M. Lukic, A. Nussbaumer, H.-P. Günther, R. Helmerich, S. Herion, M.H.
[3] Y. Zhou, Fatigue Strength Evaluation of Riveted Bridge Members, 2019. Kolstein, S. Walbridge, B. Androic, O. Dijkstra, et al., Assessment of existing steel
[4] W.M. Wilson, F.P. Thomas, Fatigue tests of riveted joints: a report, Tech. Rep, Univer- structures: recommendations for estimation of remaining fatigue life, Tech. Rep,
sity of Illinois at Urbana Champaign, College of Engineering, 1938. Joint Research Center, 2008.
[5] M. Mensinger, A. Hacker, T. Langen, Assessment of riveted railway bridges across the [9] DB-Netz, AG, Richtlinie 805–Tragsicherheit bestehender Eisenbahnbrücken, 2010.
river Pegnitz, The Eight International Conference “Bridges in Danube Basin”, [10] J.D. DiBattista, D.E. Adamson, G.L. Kulak, Fatigue strength of riveted connections, J.
Springer 2014, pp. 245–254. Struct. Eng. 124 (7) (1998) 792–797.
12 D. Leonetti et al. / Journal of Constructional Steel Research 167 (2020) 105955

[11] A.M. de Jesus, A.L. da Silva, J.A. Correia, Fatigue of riveted and bolted joints made of [23] N.G. Hansen, Fatigue tests of joints of high strength steels, J. Struct. Div. 85 (3)
puddle irona numerical approach, J. Constr. Steel Res. 102 (2014) 164–177. (1959) 51–70.
[12] A.M. de Jesus, A.L. da Silva, J.A. Correia, Fatigue of riveted and bolted joints made of [24] B. Åkesson, Fatigue life of riveted railway bridges, Ph.D. Thesis, Chalmers University
puddle ironan experimental approach, J. Constr. Steel Res. 104 (2015) 81–90. of Technology, 1994.
[13] J. Maljaars, D. Leonetti, C. Maas, Fatigue life prediction of hot riveted double covered [25] F. Baron, J. Larson, The effect of grip on the fatigue strength of riveted and bolted
butt joints, Int. J. Fatigue 124 (2019) 99–112. joints, Proc. Am. Railw. Eng. Assoc. 54 (1953) 175–190.
[14] D. Leonetti, J. Maljaars, H.B. Snijder, Fatigue life prediction of hot-riveted shear con- [26] H. van Maarschalkerwaart, Fatigue behaviour of riveted joints, Iabse Reports, vol-
nections using system reliability, Eng. Struct. 186 (2019) 471–483. ume 37, 1982, pp. 691–698 , iabse-aicp-ivbh zürich. s.
[15] R.F. Sanches, A.M. de Jesus, J.A. Correia, A. Da Silva, A. Fernandes, A probabilistic fa- [27] F. Baron, E. Larson, High-strength bolts in structural joints: a symposium: com-
tigue approach for riveted joints using Monte Carlo simulation, J. Constr. Steel Res. parison of bolted and riveted joints, Trans. Am. Soc. Civ. Eng. 120 (1955)
110 (2015) 149–162. 1322–1334.
[16] A.M. De Jesus, H. Pinto, A. Fernández-Canteli, E. Castillo, J.A. Correia, Fatigue assess- [28] A. Taras, Private Communication, to be Implemented in Assessment of Existing Steel
ment of a riveted shear splice based on a probabilistic model, Int. J. Fatigue 32 (2) Structures: Recommendations for Estimation of Remaining Fatigue Life, B. Khn, M.
(2010) 453–462. Luki, A. Nussbaumer, H.-P. Gnther, R. Helmerich Joint Research Center, 2008 In:
[17] B.M. Imam, M.K. Chryssanthopoulos, D.M. Frangopol, Fatigue system reliability anal- Tech. Rep.
ysis of riveted railway bridge connections, Struct. Infrastruct. Eng. 8 (10) (2012) [29] E. Lepretre, S. Chataigner, L. Dieng, L. Gaillet, H. Cannard, Numerical and experimen-
967–984. tal investigations of hot driven riveting process on old metal structures, Eng. Struct.
[18] J. Parola, E. Chesson, W.H. Munse, Effect of bearing pressure on fatigue strength of 127 (2016) 583–593.
riveted connections, Tech. Rep, University of Illinois at Urbana Champaign, College [30] J. Kafie-Martinez, P.B. Keating, P. Chakra-Varthy, J. Correia, A. de Jesus, Stress distri-
of Engineering, 1965. butions and crack growth in riveted lap joints fastening thick steel plates, Eng. Fail.
[19] W.H. Munse, E. Chesson, Riveted and bolted joints: net section design, J. Struct. Div. Anal. 91 (2018) 370–381.
89 (1) (1963) 107–126. [31] D.S. Simulia, Abaqus 6.12 Documentation, Providence, Rhode Island, US 261 2012.
[20] J. Carter, Fatigue in riveted and bolted single lap joints, Proc. Am. Soc. Civ. Eng. 80 [32] EN 1993-1-2, Eurocode 3: Design of Steel Structures - Part 1–2: General Rules -
(1954) 1–35. Structural Fire Design, CEN, Brussels, 2005.
[21] O. Graf, Versuche mit nietverbindungen, Springer-Verlag, 1941. [33] J. Kafie-Martinez, P.B. Keating, Finite element modeling for clamping stresses devel-
[22] B. Derby, D.A. Hills, C. Ruiz, Materials for Engineering: A Fundamental Design Ap- oped in hot-driven steel structural riveted connections, Int. J. Civil Environ. Struct.
proach, Halsted Press, 1992. Construct. Architect. Eng. 11 (5) (2017) 560–565.