ImportanteLeer EcuacionesEmpiricas

Journal of Sound and Vibration ∎ (∎∎∎∎) ∎∎∎–∎∎∎
Contents lists available at ScienceDirect
Journal of Sound and Vibration

journal homepage: www.elsevier.com/locate/jsvi
Study of the dynamic behavior of a bolted joint under heavy

loadings
Sami Daouk a, François Louf a,n, Christophe Cluzel a,b, Olivier Dorival c,d,
Laurent Champaney a, Sylvie Audebert e
a
LMT, ENS-Cachan, CNRS, Université Paris Saclay, 94235 Cachan, France
b
IUT-d'Evry Val d'Essonne, Dep. SGM, 91025 Evry, France
c
ICAM–Site de Toulouse, 31076 Toulouse, France
d
Université de Toulouse, Institut Clément Ader UMR CNRS 5312, INSA/UPS/ISAE/Mines Albi, France
e
EDF R&D, EDF Lab PARIS-SACLAY, 91120 Palaiseau, France
a r t i c l e i n f o abstract
Article history: In structural dynamics, the quantification of the quality and reliability of numerical
Received 24 March 2016 models remains a relevant issue. While the behavior of structures such as beams and
Received in revised form plates is generally understood, a large number of industrial structures are bolted assem-
22 November 2016
blies of many components with connections that are not well understood. The main
Accepted 30 December 2016
reason for this lack of understanding is that the dynamical behavior of the whole as-
Handling Editor: Prof Z. Su
sembly depends critically upon joint conditions, especially under heavy loadings. For the
purpose of improving our understanding and the development of pertinent models, a
Keywords: dynamic test bed, based on a bolted structure, is designed and modal testing is performed.
Structural dynamics
The configuration of the bolted joint and the level of the loading are the relevant para-
Bolted joint
meters, related to joint conditions, that are considered in this study. The results of the
Modal testing
Specific damping capacity experimental campaign show the variation of the dissipation in a bolted joint and its
Dissipation apparent stiffness as a function of joint conditions.
& 2017 Elsevier Ltd All rights reserved.
Introduction
The dynamical behavior of industrial assemblies has to be reliably understood in order to avoid accidents that would
endanger the safety of people and facilities. In particular, the dynamic behavior of mechanical systems in nuclear plants has
to be predicted with full confidence. The stiffness of these assembled structures results from keeping relatively flexible
components in position with many fasteners such as preloaded bolts. These fasteners contribute to the distributions of
stiffnesses and masses in the assembly in a complicated way, making the system nonlinear and generating uncertainties and
variabilities [22,15].
Many difficulties encountered when analyzing nonlinear systems come from a lack of understanding of the physical
nonlinear phenomena. As a result, small modifications in modeling nonlinear elements can lead to substantial discrepancies
between the responses of the system [6,7]. In fact, even though the sources of uncertainties can be identified, the char-
acterization of their impact on the dynamic behavior of the system is still not clear. Therefore, when studying the dynamics
n
Corresponding author.
E-mail address: francois.louf@lmt.ens-cachan.fr (F. Louf).
http://dx.doi.org/10.1016/j.jsv.2016.12.047
0022-460X/& 2017 Elsevier Ltd All rights reserved.
Please cite this article as: S. Daouk, et al., Study of the dynamic behavior of a bolted joint under heavy loadings, Journal of
Sound and Vibration (2017), http://dx.doi.org/10.1016/j.jsv.2016.12.047i
2 S. Daouk et al. / Journal of Sound and Vibration ∎ (∎∎∎∎) ∎∎∎–∎∎∎
of bolted joints, the lack of a deep complete comprehension of what happens in the contact zones raises prominent
questions:
How are the damping and stiffness of the system influenced by the presence of bolted joints ?
How do the damping and stiffness of the system vary depending on the excitation level of the system ?
From a fundamental viewpoint, the effects linked to the contact of two surfaces on the dynamic behavior are the subject
of many experimental research activities. For instance, the effects of normal loadings are studied in [2] and the effects of
shear uniform slip are presented in [30]. Naturally, modeling these friction problems from the description of micro-slips and
their impact on damping are also the subject of numerous analyses. In [32], different bolted joints of thin plates are studied
and the evolution of dissipation as a function of the frequency, the amplitude and the mode are presented. In [21], we see
that running-in fatigue effects of surfaces have an impact on damping functions. In [16,22], the bolt preload is the added
parameter. We find in [18] a review of models and applications on different types of bolted joints. Two conclusions are that
the choices of models depend on the application type and that the variations of friction properties and of contact forces
distributions generate high dispersions of joint stiffness and damping. Simulations of nonlinear friction contact effects of
two bolted plates under shearing are presented in [29] via a local finite elements contact analysis and a more global
modeling of the joint. However, in this study too, a conclusion is that the variability of the structure responses forces to
experimentally update the parameters on a joint of the same type as the one used.
In addition, many authors analyse joints directly via tests on mini-structures using a particular technology, then exploit
the results to simulate the effects of small joint modifications or understand the influence of loadings on dissipation. Hence,
we find in [9,24] an experimental identification of dissipations in bolted joints of space launcher structures (SYLDA). Then, in
[10], a numerical exploitation is carried out after updating friction parameters, including their fatigue evolutions. In [28], all
components of a bolted and glued joint between two satellite sandwich panels are studied experimentally. The authors of
[20] introduce the effects of viscoelastic damping of elastomer layers in a structural aircraft joint.
Moreover, in relation to joint conditions, the study of a two-beam stainless steel structure coupled by three bolts [8]
shows that the presence of a bolted joint has a significant impact on the frequency response functions of the system. In
particular, the results of this study showed that the torsional modes are sensitive to the interfacial conditions and that the
bending modes are sensitive to the bolt preloads and the order in which the bolts are tightened. In order to distinguish the
effect of the mechanical joint from other sources of uncertainty and nonlinearity, a second study on the same structure was
conducted [31]. Its focus is on the effects of different loading conditions, measurement techniques and boundary conditions
on the response of the system. Another experimental study has been recently performed on a portal frame containing
elementary steel-aluminum bolted joints [33]. The results show that input parameters such as the bolt preload and the
tightening order have a notable impact on some natural frequencies but the loading level was low and not controlled.
The vast majority of these articles describes interfaces between two bolted planes subject to local shear stresses. When
joints are thin and are subject to transverse moments or loadings, the levels of interfacial stresses are quickly significant and
generate micro-slips. Very few articles study massive structures having a joint plan normal to the average direction of the
structure. The structure used as a reference in this article is a large booster bump, with a massive body, large diameter joints
and a rotating axis. A great part of the bolted assemblies used in rotating machinery are subject to high loading and
vibration levels. The understanding of their dynamical behavior under such conditions turns out to be crucial during their
design. However, most of the research studies conducted in this field consider joints comprising few bolts with low preloads
and low excitation levels. In order to generate significant slips, the loading level of such joints has to be very high which
creates experimental difficulties if precise measurements are requested. In addition, analyzing a symmetric mode means
that slips cannot occur in the joint faces. The usual joint place slip models are then baseless and experimental data is
necessary. This is the purpose of this article, leading thereafter to the estimation of uncertainties on prediction of damping
and natural frequencies. Consequently, we show the design of a dynamic test bed containing a configurable bolted joint,
then consider heavy loadings and examine the experimental variability due to the variation of joint conditions.
The first part of this paper describes the specifications, the basic ideas and the iterative sizing procedure of this test bed.
The second part presents the experimental procedure and the first results of modal testing of the assembly. In the third part,
global experimental results (variation of dissipation and natural frequencies) obtained on the first bending mode of the
assembly are presented and discussed. The fourth part presents the exploitation of experimental results on the level of the
joint of interest in order to describe the evolution of the dissipation in the joint and its apparent stiffness depending on joint
conditions.
1. Design of the assembly
1.1. Context of the study and related technical specifications
The work presented in this paper is carried out in partnership with EDF R&D and thus should be representative of its
industrial assemblies such as booster pumps used in thermal units. These assemblies comprise bolted joints having the
following features: the bolted parts are metallic, massive, the joint is plane and has a large diameter containing many bolts
S. Daouk et al. / Journal of Sound and Vibration ∎ (∎∎∎∎) ∎∎∎–∎∎∎ 3
Fig. 1. Illustration of a one-stage booster pump with several bolted joints (bolts are not represented).
(see Fig. 1). Depending on the modes of vibration, the joints of interest can behave in several ways: bending, shearing,
torsion and tension. The frequencies considered are below 1000 Hz.
For the purpose of improving our understanding and the development of pertinent models, it is essential to quantify
experimentally the impact of parameters associated with similar bolted joints on the variation of the dissipation and
stiffness of the joint [16]. The most relevant parameters that we want to vary are:
the configuration of the bolted joint: bold preload [27,23], number and position of preloaded bolts [34];
the type and level of the loading applied to the bolted joint [8,31].
1.2. The bolted joint
The bolted joint that will be tested must have similar characteristics to the ones found in assemblies used in thermal
units (see Fig. 1). However, for reasons of size and space, we will design a joint of smaller size (diameter of circa 20 cm) with
M12 bolts.
1.3. Operating principle
The system is preloaded by tightening bolted parts, which generates static loads in the joint. During dynamic excitation
of the system, dynamic loads will overlap with these static loads. If variations of dynamic loads are small compared to static
loads, then there will be little impact on the dynamic response in terms of dissipation and natural frequency. That agrees
with the threshold effect of disbonding or slip mechanisms described in [28,10]. It is therefore important to be able to
control the loading level in the joint and to make it significant compared to the static loads coming from tightening the
bolted parts.
Nevertheless, the application of high loading levels in dynamics is not possible with an experimental material of rea-
sonable size. One interesting solution is to take advantage of the dynamic load amplification at resonance. This way, when
injecting little energy in the system, the amplitude of the dynamic response can be significant. In addition, it is important to
ensure that the first modes heavily load the bolted joint of interest and that they are associated with low frequencies.
Therefore, two offset masses are connected to the joint by beam type elastic resonators. The conversion of kinetic energy of
the masses into elastic energy in these elements enables the transfer of the desired loads to the central bolted joint.
Furthermore, given that the assembly comprises many joints, it is essential to be provided with a means of extracting all
information inherent in the bolted joint of interest. A practical way of achieving that is to design a non dissipative massive
part, that we call “reference part”, having similar stiffness and inertia to the joint. The contact interfaces between the beams
and the bolted joint or between the beams and the reference part must behave similarly, particularly in terms of dissipation.
Thus, these joints have identical geometries, flatness requirements and contact surface roughness. In addition, this part was
designed in a way that its mass can be modified by adding setscrews, in order to match different configurations of the bolted
joint. Thus, replacing the central bolted joint by this part leads to the identification of the dissipation in the joint and the
stiffness variations of the joint, through a differential analysis. Applications of this approach are found in [28,14,11] where it
is used to identify friction and dissipation in different mechanical components and joints.
In short, the design of the assembly is based on three principles:
load the assembly at resonance in order to reduce the energy to inject, even for high loading levels;
carry out a differential analysis so that the dissipation in the bolted joint of interest can be estimated;
Fig. 2. Geometry of the assembly with a central bolted joint (J0 ) and of the reference assembly; there are other bolted joints (J ) and (J ′) in both
configurations.
use a beam type resonator to measure the loads in the joint.
The preliminary designs of the assemblies are shown in Fig. 2. The system, having a mass around 40 kg, can be easily
disassembled so that masses can be removed or added, and their distance to the joint can be modified without affecting the
joint conditions.
1.4. Sizing procedure
The main criteria for sizing the test bed are, in decreasing order of importance:
the loading level must ensure that slips and/or gaps of a few μm occur in the contact zones between the parts of the
bolted joint or between bolts and flanges; these displacements will influence the dissipated energy and the value of the
eigenfrequencies;
even when highly loaded, the whole assembly must have an elastic behavior;
the first natural modes have to be clearly separated, hence associated with frequencies below 1000 Hz.
In order to ensure an optimal sizing of the joint and other parts, we sought to modify the following parameters during
the iterative sizing procedure:
The geometry of the flanges of the bolted joint of interest (diameter, length, thickness of the joint faces, etc.)
The length and the diameter of the elastic beams;
The geometry and the position of the masses.
For each mode under study, an iteration of the sizing procedure comprises the following basic steps:
look for the shape of the natural mode and the natural frequency associated with it depending on the chosen geometry
and the additional masses: this step is quick because an analytical model had been built;
for a loading of the joint compatible with the shape of the mode under study (moment, shear force, …), look for the load
amplitude generating satisfactory local gaps and slips; for this purpose, a quasi-static nonlinear finite element analysis
(contact, friction) is carried out on the joint (see Fig. 3) where a shear force T and a bending moment Mf are applied; the
stress levels in the joint and the other parts of the assembly are also checked at this stage;
for a given load at resonance, while assuming a maximal modal damping ratio ζ = 0.2 % , proceed to an analytical estimate
of the link between the amplitude of the load in the joint and the force to apply; this step helps verify that the minimal
capacity of the shaker is sufficient.
The final geometries of the bolted joint and the reference part are shown in Figs. 4(a) and (b) respectively.
Fig. 3. The structural model used to compute the quantities of interest in the joint faces – Loading applied: T ¼13000 N and Mf ¼ 6500 N.m.
Fig. 4. Geometries of the tested bolted joint and of the reference part, having similar mass and moments of inertia.
2. Experimental setup
2.1. Experimental evaluation of dissipation
2.1.1. Quality factor

A first approach to experimentally evaluate the dissipation inherent in the presence of a bolted joint in a structure may
consist of using the modal damping ratios ζ given by modal testing softwares.
The evaluation of these ratios can be done by using the measuring method at 3 dB [17]. That means considering the
angular frequencies Ω1 and Ω2 at 3 dB from the maximum of amplitude of the transfer function at resonance angular
frequency Ωr (see Fig. 5). It is also conventional to use the quality factor Q which is the ratio of the resonance angular
frequency and the bandwidth at 3 dB:
Ωr Ωr
Q= =
ΔΩ Ω2 − Ω1 (1)
In case of small damping, an approximation of the modal damping ratio ζ is:

1
ζ≈
2Q (2)
This measure assumes that damping is constant depending on the level of vibrations within the 3 dB bandwidth,
which makes it precise solely in case of a small damping variations against the loading level.
2.1.2. Specific damping capacity (SDC)

An alternative method is to measure the specific damping capacity (SDC) ψ, which intends to measure damping for a
given deformed shape such as a modal shape. This method was used to assess damping in composite sandwich panels
[1,25,26,28] and in electric cables of a satellite structure [11]. It is expressed as a ratio of the energy dissipated per cycle and
the elastic (or stored) energy per cycle:
Wd
ψ=
We (3)
The coefficient ψ is linked to the modal damping ratio ζ, when the latter is small, as follows:
ψ = 4πζ (4)
In case of forced oscillations, the dissipated energy Wd is calculated by summing the power injected by the shaker in the
assembly per cycle as:
T → → T → →
Wd = ∫0 F . v dt = ∫0 F . du
(5)
Particularly in our assembly, the elastic energy We is mostly located in the elastic beams connecting the masses to the
joint. In case of the first symmetric bending mode, the shape of the bending moment in a beam can be calculated using the
model presented in Fig. 6. Since the mode is symmetric, the rotation at x ¼0 is null. The mass M0 represents half of the mass
of the joint. The offset mass at one end of the beam is represented by its mass ML and moment of inertia IL.
E, ρ, I, et S are respectively the Young's modulus, the density, the second area bending moment, and the cross-section of
the beam. As shown in [19], the general shape V(x) of a bending mode of a beam is:
V (x) = A cos (βx) + B sin (βx) + C cosh (βx) + D sinh (βx) (6)
with
ρS
β 4 = ω2
EI (7)
and ω the angular natural frequency of the mode. The condition of null rotation at x ¼0 and the equilibrium conditions of
the masses at x ¼0 and x ¼L give the following relations:
Fig. 5. Illustration of the 3 dB bandwidth.
Fig. 6. Simplified analytical beam model used to estimate the frequency of the first symmetric bending mode.
Fig. 7. Shape of the displacement, bending moment and shear force of the beam model in Fig. 6, in the case of the first bending mode.
β4
V ′(0) = 0 V ‴(0) = M0 V (0)
ρS (8)
β4 β4
V ″(L ) = IL V ′(L ) V ‴(L ) = − ML V (L )
ρS ρS (9)
that lead to nonlinear equation of β. Its first root gives the first natural frequency and first natural mode of the structure. The
bending moment Mf and shear force T can be calculated from the obtained mode V(x):
Mf (x) = EIV ″(x) T (x) = − EIV ‴(x) (10)
For a geometry and materials representative of the designed test bed, the shapes of V(x), Mf and T are plotted in Fig. 7 for
Mf (0) = 1000 N. m .
It is clear that the bending moment variation is nearly linear. That implies that it can be accurately estimated on the
whole beam if estimated at two points. Therefore, when arranging strain gauges in two half-bridges on each beam, we are
able to assess the curvature at two points. After static calibration, we can estimate the bending moment at these two points
and thus the expression of the moment M f (x ) on all the beam.
Then, the elastic energy is given by the relation:
f (x)2
M
We = 2 ∫Beam 2 EI
dx
(11)
For the sake of simplicity, we considered the shape of the bending moment to be linear which leads to an error in the
evaluation of the elastic energy We that is below 2 %.
When using this energy based method, no assumption is made about the damping variation as a function of the level of
vibrations. Thus, this measure leads to a precise characterization of the damping of each mode for any loading level in the joint.
2.2. Instrumentation
In this work, we focus on the first bending mode only since it is associated with a high dynamic amplification factor.
Several experimental tools are needed in order to measure accurately the quantities of interest. In order to excite this mode,
→
a shaker is placed radially at the end of the assembly as presented in Fig. 8(a). The force F is measured with a force sensor
installed on the excitation rod. Uniaxial and triaxial accelerometers were used to measure the frequency response function
of the assembly, and hence its natural frequencies. As shown in Fig. 8(a), an accelerometer installed in front of the shaker
→
leads to the velocity v through signal integration.
Given that the level of loading applied to the bolted joint is an important parameter to vary, it is essential to have a
means of measuring it accurately. Therefore, sixteen strain gages were glued on the beams as shown in Fig. 8(b). They were
arranged in eight half-bridges (four in the horizontal plane and four in the vertical plane) to measure bending strains.
2.3. Experimental procedure
The experimental procedure is structured into three parts. First, it is necessary to carry out a modal analysis of the bolted
assembly in order to estimate its natural frequencies, especially the frequency of the normal mode of interest: the first
symmetric bending mode. Free-free boundary conditions are applied so as to limit the number of joints.
After this first step, a sine sweep is performed near the natural frequency of the first bending mode. During this sweep,
the force injected by the shaker is automatically adapted in order to generate the desired value of the moment in the joint.
This step is accomplished through the “Tracked Sine Dwell” application of the software LMS Test.Lab, and leads to:
the link between the desired moment in the joint and the amplitude of the applied load, and
the resonance frequency for each value of the moment in the joint.
The last step is to apply, for a given value of the moment in the joint, a sinusoidal force at the resonance frequency
previously measured. During this test, the time evolution of the excitation force, the acceleration at the excitation point and
the strains are measured. These measures are then used to calcule the coefficient ψ as described in Section 2.1. The last two
steps are carried out for each desired value of the moment in the joint, and for each tested configuration of the bolted joint.
In addition to these tests on the bolted joint, the same procedure is carried out on the reference assembly (steps 1, 2 and 3)
for the same values of the moment in the central part of the assembly.
2.4. Experimental modal analysis of the assembly
The forced response of the structure with the central bolted joint to a white noise excitation in the frequency range (0–
1024) Hz is shown in Fig. 9.
Fig. 10 shows the first 4 bending modes found. Due to misalignment and mispositioning of the excitation stringer, other
modes can be noticed in Fig. 9. In ascending order, the frequencies of the bending modes below 1000 Hz are:
77.7 Hz: first symmetric bending mode;

276.1 Hz: first antisymmetric bending mode;
557.4 Hz: second symmetric bending mode;
829.0 Hz: second antisymmetric bending mode.
Fig. 8. The different sensors, installed on the test bed, in order to measure the injected energy, elastic energy and the level of the loading applied to the
bolted joint.
Fig. 9. Frequency response function of the structure with bolted joint subject to a white noise excitation, specifying the first 4 bending modes.
→
Fig. 10. The first 4 bending modes of the first configuration tested: projection on the y direction.
3. Results on the first bending mode
3.1. The various configurations tested
Several configurations of the test bed were tested at the first bending mode:
the reference configuration (i.e. with the reference part);

the joint with 16 bolts tightened at 70 N.m;
We sought to consider two distinct bolt preloads, namely 70 N.m and 40 N.m. The M12 bolts used have a diameter of 12 mm
and a grade of 12.9. The elastic limit of the screws is: σe = 12 × 9 × 10 = 1080 MPa . The Von Mises stress in the screw is
given by the normal stress and the shear stress that results from torsion:
σ VM = σ 2 + 3τ 2 (12)
with
4F0 32C0 d 16C0
σ= et τ = =
πd 2 πd 4 2 πd 3 (13)
and
C0
F0 =
0.16p + μ10.582d2 + μ2 rm (14)
where d = 12 mm , p = 1.75 mm , d2 = d − 0.6495p = 10.86 mm and rm = 7.5 mm , for M12 screws. For the sake of simplifi-
cation, we consider that μ1 = μ2 = μ in the following. For a minimal friction coefficient μ = 0.1, a torque C0 = 70 N. m leads to
a preload F0 = 42000 N and to a Von Mises stress σ VM = 515 MPa = 0.5 × σe . For μ = 0.1, a torque C0 = 40 N. m leads to a
preload F0 = 24000 N and to a Von Mises stress σ VM = 295 MPa = 0.3 × σe .
For all these configurations, the values of the moment applied to the joint varied between 250 N. m and 2250 N. m . These
values of the moment in the joint can be qualified as “heavy loading” for three main reasons. The first one derives from the
dynamic response of the assembly subject to a white noise excitation. The time signal on the strain gages showed a maximal
axial strain of ε = 2.3 × 10−6 , leading to bending moment Mf = 4.3 N. m at the measuring section. Assuming that this strain
corresponds to the first bending mode, the value of the moment in the central bolted joint is thus Mfc = 6.7 N. m .
Let us then quantify the variations of forces in the fastening elements, using equation (14). A finite element computation
performed on a flange helps estimate the variation of forces in each screw as a function of the moment applied to the joint.
For a moment of 2500 N. m in the joint, with μ = 0.1 or μ = 0.2 and bolt preloads of 70 N. m and 40 N. m , variations of the
force in the most loaded screw can vary more than 10 %.
The last reason comes from measuring the variation of contact pressure between the preloaded configuration and the
one loaded with a given moment. Therefore, we calculated the integral of the contact pressure in the zone z > 0 and in the
zone z < 0 of the preloaded configuration and the one loaded with a moment equal to 2500 N. m . For the same friction
coefficients and bolt preloads, the mean contact pressure in the upper and lower zones of a flange vary between 10 % and 20
%, and sometimes over 30 %. These calculations show that the loading applied to the joint significantly changes the pressure
levels in the different contact zones.
Our aim is to estimate the influence of the configuration of the central bolted joint (J0 ) on the energy dissipated at a
vibration mode. The system produced, whose preliminary design is shown in Fig. 2(a), comprises other bolted joints marked
(J ) and (J′). These joints will naturally dissipate energy too. In order to be able to isolate the energy dissipated solely in the
joint (J0 ), we focused on two points:
the energy dissipated in joints (J ) and (J′) must not be too large in comparison with the energy dissipated in the joint (J0 ). If
so, the latter could be considered as noise. Therefore, the parts forming these joints are very massive (thick flanges), thus
very stiff, and are highly preloaded at 110 N. m . This torque value is reasonable for the M12 bolts of grade 12.9 used. For a
minimal friction coefficient μ = 0.1, this leads to a preload equal to 66000 N and to a Von Mises stress
σ VM = 810 MPa = 0.75 × σe .
Despite the careful design of joints (J ) and (J′), the energy dissipated in them cannot be nil. Therefore, the reference
system, whose preliminary design is shown in Fig. 2(b), comprising the same joints (J ) and (J ′), was produced. Using a
reference system added to measuring the moment in the joint at each instant enables to subtract the energy dissipated in
joints (J ) and (J ′) from the energy dissipated in the system shown in Fig. 2(a), and have the energy dissipated solely in the
joint (J0 ). This approach will be referred to as “differential analysis” hereinafter.
Finally, since we study the first symmetric bending mode, no slips can occur on the interface of the joint (J0 ). Therefore, in
joint (J0 ), the only slips that can occur will be in the contact zones between the screw heads or the nuts and the flanges. We
can thus expect to have a low level of energy dissipated at this first mode.
3.2. Evolution of ψ and the resonance frequency
Fig. 11 shows the evolutions of ψ and the resonance frequency for the two extreme configurations tested with the central
bolted joint.
Firstly, we notice that the resonance frequency decreases regularly with the amplitude of the moment in the joint. This
confirms that the behavior of the system is nonlinear, and that local disbondings appear and increase the flexibility of the
structure when the moment in the joint increases.
Table 1 shows the relative variations of the resonance frequency for these configurations (see the first two rows of the
table). The variation is low in both cases. However, the variation is higher with 8 bolts tightened at 40 N.m than with 16
bolts tightened at 70 N.m; This is an expected result. Indeed the frequencies vary slightly but that does not mean that the
apparent stiffness of the bolted joint changes very little. In fact, half of the system located on a side of the plane of symmetry
can be modelled by two springs configured in series: one spring representing the angular stiffness of the bolted joint kJ, and
another representing the angular stiffness of the beam kB. The stiffness equivalent to both elements is:
1
k eq =
1 1
+ kB
kJ (15)
The order of magnitude of the angular stiffness of the bolted joint is kJ ≈ 100 × 105 N. m. rad−1. The order of magnitude of
the angular stiffness of the beam loaded at one end is 1.8 × 105 N. m. rad−1. So there is a factor exceeding 50 between both
stiffnesses. Therefore, a 50% relative variation of the stiffness kJ leads to a relative variation of the equivalent stiffness of:
Fig. 11. Evolutions of ψ and the resonance frequency with the amplitude of the moment in the bolted joint for two extreme configurations.
Table 1
Relative variations of the resonance frequency in the different configurations tested.
Config. Low moment High moment Relative

variation
8 bolts - 40 N.m 78.08 Hz 77.44 Hz 0.83 %

16 bolts - 70 N.m 77.56 Hz 77.25 Hz 0.40 %
Reference 79.04 Hz 78.89 Hz 0.19 %
1 1
+
Δk eq kJ kB
= − 1 = − 1.74%
k eq 1 1
+
0.5kJ kB (16)
The small variation of the resonance frequency does not mean at all a small variation of the stiffness of the joint. It is due
to a big stiffness difference between the beams and the central part. This leads to a relative frequency variation of −0.87% .
For maximal loading, the coefficient ψ found for the different configurations is in close to 14 × 10−3. This value matches
with a modal damping ratio of 0.11%. The dissipations in the designed system are then very small for the first bending mode.
Secondly, we notice that the coefficient ψ evaluated from the quality factor Q and the one calculated from the measure of the
energy injected in the assembly are slightly different. The former is always below the latter. Fig. 12 helps explain this point.
For instance, let us consider a moment M = 1858 N. m , for which the SDC measured with the energy based method is
ψ = 0.0116 J . Let A (M, ψ ) be this starting point, that is a measured one. The strain amplitude, thus the amplitude of the
moment, at −3 dB is M−3 dB = 1858/ 2 = 1314 N. m . Let B be the point having coordinates (M−3 dB, ψ ). When using the ex-
perimental curve again, we find that a moment M−3 dB = 1314 N. m should lead to a SDC of ψ−3 dB = 0.0084 . We arrive then to
the point C (M−3 dB, ψ−3 dB ). Therefore, if the SDC chosen for an amplitude of moment M is the one estimated for an amplitude
M 3 dB, we would get the point D having coordinates (M, ψ−3 dB ). In practice, the classical method at −3 dB is equivalent to
considering that the SDC is constant within the range [M−3 dB M ] and thus choose a mean value of the SDC between ψ and
ψ−3 dB . It makes then sense to find the corresponding curve between the curves of the energy based method, namely te one
measured and its offset.
3.3. Evolution of ψ and the resonance frequency for the reference configuration
Fig. 13 shows the evolutions of ψ and the resonance frequency for the reference configuration.
The results achieved lead to comments similar to the ones in Section 3.2. We notice however that the coefficient ψ is
smaller than in the configurations with the central bolted joint. This is an expected result since a part of the sources of
dissipation was removed in this assembly. The resonance frequency has again a regular but slight variation. Table 1 shows
that the relative variation is of 0.19 % only. It is then below the values obtained in the configurations with bolted joint that
were studied. This result is expected because a part of the nonlinear contact phenomena cannot occur with the reference
part.
Fig. 12. Interpretation of the gap between the values of ψ obtained through the energy based method and the 3 dB method.
Fig. 13. Evolutions of ψ and the resonance frequency with the amplitude of the moment in the joint for the reference configuration.
Fig. 14. Two-stage updating strategy applied to estimate the apparent stiffness of the central bolted joint as a function of joint conditions.
4. Exploitation of experimental results through differential analyses
Since the reference configuration shows variations of the resonance frequency and of the SDC with respect to the loading
level, it is clear that a differential analysis is required to extract the effects of the bolted joint solely. The differential analysis
aims at removing the effect of the other joints in the system, assuming that their effects, for a given loading level, are the
same for the reference configuration and for the bolted one. Therefore, a two-stage updating process is carried out. The
masses of the central bolted joint and reference part being the same (for either 8 or 16 bolts), the frequency variation is
assumed to be related to a variation of the joint stiffness. In the first step of the differential analysis (Section 4.1), the joint
stiffness is identified. In the second step (Section 4.2), the evolution of the SDC solely in the bolted joint as a function of the
joint conditions is described.
4.1. Evolution of the apparent stiffness of the bolted joint
Fig. 11(c) and (d) show significant variations of the resonance frequency in presence of the central bolted joint. This
indicates that nonlinear mechanisms appear in the joint when vibrating. However, Fig. 13(b) shows the evolution of the
resonance frequency of the reference configuration. The frequency variations imply that nonlinear mechanisms exist in this
system too, due to the bolted joints connecting the beams to the other parts. Therefore, it is not possible to assign the
nonlinear behavior observed to the central bolted joint solely. A differential analysis is necessary in order to isolate the
nonlinearities related to this joint solely.
An updating strategy, presented in Fig. 14, is then suggested. The first stage is to work on the reference configuration and
update a stiffness parameter related to the beams (elastic modulus) that represents the global stiffness variation due to
nonlinearities in all the other joints. The shape of the natural mode studied, being the same in the reference and bolted
configurations, we consider that, for a given moment, the nonlinear effects in the other joints are the same in both con-
figurations. Nevertheless, in case of a bolted configuration, nonlinearities are present in the central bolted joint. The second
stage of the updating strategy is then to update a stiffness parameter related to this joint, for a given elastic modulus of the
beams provided by the first stage.
The two-stage updating strategy uses a simple numerical model (see Fig. 15) in order to estimate the frequency of the
first symmetric bending mode in different configurations:
the flexibility of the central part, massive or bolted joint, is modeled by a spring of angular stiffness k;
the resonator connecting the offset masses to the central part is modeled by an elastic Euler-Bernoulli beam of elastic
modulus E;
Fig. 15. Model used to estimate the frequency of the first bending mode for different configurations.
the center of gravity of the offset masses is rigidly connected to one end of the beam;
the elastic beam is rigidly connected to the anchor point of the spring.
In this model, the mass parameters can vary because they depend on the configuration under study, but they are known.
The first symmetric natural mode v(x) is sought in the basis (φ1, φ2, φ3 )
3
v (x) = ∑ γi φi (x)
p=1 (17)
with
φ1 (x) = 1 (18)
φ2 (x) = x (19)
L ⎛ x − λ ⎞⎟
φ3 (x) = (H (x − λ ) − H (x − λ − L )) (x − λ )2 ⎜ 1 −
2 ⎝ 3L ⎠ (20)
→
where H is the Heaviside step function. The function φ1 represents the translational rigid body mode along y , and φ2
→
represents the rotation of the structure around z that is possible because the angular stiffness k is not infinite. The function
φ3 represents the deformation of the flexible beam. It corresponds to the deformed shape of a beam solicited by a point
force at end. This choice was guided by the presence of significant masses at the end of the beam.
The coordinates of the first mode in the basis (φ1, φ2, φ3 ) and the natural frequency associated with it are found when
considering that the Rayleigh-Ritz ratio is stationary, then solving the 3 3 eigenvalue problem that results. This model is
used to estimate the natural frequency of the first symmetric bending mode f (E, k ), for a given mass distribution depending
on the analyzed configuration.
As a first stage, the resonance frequencies of the reference assembly are exploited. The stiffness k = k 0 of the central rigid
part is considered known and fixed: it is identified through a simple linear static finite element calculation. The boundary
conditions are specified in Fig. 16(a) and the norm of the displacement field obtained with a fine mesh of second order is
shown in Fig. 16(c)). For every moment Mi, we sought the elastic modulus of the beams that minimises the difference
between the model response fi (E, k = k 0 ) and the measured resonance frequency fi, mes . For this purpose, the difference:
(
ei2 = fi (E, k = k 0 ) − fi, mes )2 (21)
is minimized through a gradient descent with a variable step size:
Ei = argmin ⎡⎣ fi (E, k = k 0 ) − fi, mes ⎤⎦

2
E> 0 (22)
The modulus E0 = 2.0 × 1011 N. m−2

is chosen as initial value of E during the updating process. At the end of this first
updating stage, a set of couples (Mi, Ei ), shown in Fig. 17(a), are achieved leading to a good correlation between the model
and the measurements in the reference configuration.
As a second stage, for each bolted configuration tested and each applied moment Mi, we sought the joint angular stiffness
k that minimises the difference between the model response fi (E = Ei, k ) and the measured resonance frequency fi, mes :
ki = argmin ⎡⎣ fi (E = Ei, k ) − fi, mes ⎤⎦

2
k>0 (23)
At the end of this second updating stage, and for each bolted configuration tested, a set of couples (Mi, k i ), shown in Fig. 17
(b), are achieved leading to a good correlation between the model and the measurements in bolted configurations.
Fig. 17(b) highlights the nonlinear behavior of the bolted assembly. For all bolted configurations tested, the joint stiffness
drops by 30% when the moment varies from 250 N. m to 2250 N. m . In addition, we see that the configuration with 16 bolts
is stiffer than the configuration with 8 bolts, as expected. Finally, in this specific case study, the effect of the bolt preload on
Fig. 16. Linear numerical models used to estimate the stiffness k0 of the reference part in the first updating and then initialize the stiffness of the bolted
joint stiffness in the second updating.
the apparent stiffness of the bolted joint seems to be much less preponderant than the loading level and the number of
bolts, for a fixed number of preloaded bolts. We notice however that the angular stiffness k decreases, when the moment in
the joint increases. This effect is more important in the case of low preloads than in the case of high preloads.
4.2. Evolution of the dissipation in the bolted joint
For each bolted configuration tested, and for each value of the moment applied to the joint, the energy dissipated in the
reference assembly is subtracted from the energy dissipated in the bolted assembly. That leads then to the evolutions of the
energy dissipated solely in the joint, for each configuration, as a function of joint conditions. The results are presented in Fig. 18.
The dissipated energy increases with the amplitude of the vibrations, which is in proportion to the amplitude of the
moment in the joint. If damping was independent of the loading level, the growth would be in proportion to the square
value of the moment. However, in practice, applying a power regression law to the results shows that the exponent is clearly
different than m ¼2. This confirms that the behavior is nonlinear and that the loading level has an impact on the dissipative
phenomena in the joint.
Fig. 17. Results of the two-stage updating strategy presented in Fig. 14. (a) Evolution of the normalized elastic modulus E of the beams obtained at the end
of the FMU on the reference configuration. (b) Evolution of the joint angular stiffness k (N.m.rad 1) obtained at the end of the SMU, for each bolted
configuration.
Fig. 18. Evolutions of the energy dissipated in the central bolted joint for each bolted configuration tested.
Fig. 19. Evolution of the dissipation in the bolted joint for each bolted configuration tested.
Extracting the energy dissipated solely in the joint enables us to build a coefficient ψJ denoting the contribution of the
bolted joint to the dissipation. For a given loading level, ψJ is calculated as the ratio of the dissipated energy in the joint
solely and the maximal elastic energy on the whole assembly. Fig. 19 presents the evolution of ψJ as a function of the loading
level. It shows the variation of ψJ for each joint configuration and compares it to ψ that is calculated including all the
dissipations in the assembly. This first analysis shows that the dissipation due to the joint is relatively low compared to the
whole experimental assembly, namely around 1/4.
When zooming in, we see that for three of the four bolted configurations tested, ψJ varies very little until a certain
loading level and then increases. When the moment in the joint is small, slip zones change slightly. Then, when the loading
exceeds a certain level, slip zones start changing and sources of dissipation become relevant.
5. Conclusions and future work
The first part of the paper described the iterative sizing procedure of a dynamic test bed, based on a bolted structure,
where joint conditions and loading levels may vary. A configurable bolted joint and a non dissipative part with similar
characteristics were designed. The properties of the structure, for a given bolted joint, were optimized in order to satisfy a
set of conditions in terms of slips, gaps and maximum stress level. The second part showed different methods used to
evaluate local dissipation experimentally and presented the experimental procedure. The last part exhibited the results of
tests performed at the first symmetric bending mode of the structure. They lead to the conclusion that the loading level has
an impact on the resonance frequency and modal damping. Through differential analyses, we found that the stiffness of the
bolted joint plummets when the loading level increases. The evolution of the energy dissipated solely in the joint revealed
the existence of a threshold, below which the joint has a significantly linear behavior, and beyond which slip zones evolve
and dissipations increase.
The tests performed at the first symmetric bending mode show that dissipations in the bolted joint of interest are small
and that the configuration of the bolted joint has low impact on the dynamic behavior of the whole structure. Current tests
at the first torsional mode, which would generate larger slips in the bolted joint, might reveal a higher dependency of
dissipation on the bolted configuration being tested. Alternatively, a practical solution could be to generate shearing in the
bolted joint at the first symmetric bending mode by breaking the symmetry of the whole assembly.
In the framework of the SICODYN research project [3,4], initiated in 2012 and carried out till 2016, this experimental
modal analysis will be extended to the case of a bolted joint connecting two parts of a one-stage booster pump used by EDF
company in its thermal units. That would lead to a better understanding of experimental variability and then to an as-
sessment of the ability of parametric and nonparametric probabilistic numerical methods [12,5,13] to propagate and
quantify uncertainties in bolted assemblies.
Acknowledgements
This work has been carried out in the context of the FUI 2012–2016 SICODYN Project (SImulations credibility via test-
analysis COrrelation and uncertainty quantification in structural DYNamics). The authors would like to gratefully ac-
knowledge the support of the FUI (Fonds Unique Interministériel).
References
[1] R. Adams, M. Maheri, The dynamic shear properties of structural honeycomb materials, Compos. Sci. Technol. 47 (1) (1993) 15–23.
[2] C. Andrew, J. Cockburn, A. Waring, Metal surfaces in contact under normal forces: Some dynamic stiffness and damping characteristics. in: Pro-
ceedings of the Institution of Mechanical Engineers, Conference, vol. 182, SAGE Publications, 1967, pp. 92–100.
[3] S. Audebert, SICODYN International benchmark on dynamic analysis of structure assemblies: variability and numerical-experimental623 correlation
on an industrial pump, Mech. Ind. 11 (6) (2010) 439–451.
[4] S. Audebert, A. Mikchevitch, I. Zentner, SICODYN international benchmark on dynamic analysis of structure assemblies: variability and numerical-
experimental correlation on an industrial pump (part 2), Mech. Ind. 15 (1) (2014) 1–17.
[5] A. Batou, C. Soize, S. Audebert, Model identification in computational stochastic dynamics using experimental modal data, Mech. Syst. Signal Process.
50–51 (0) (2015) 307–322.
[6] M.R. Brake, The effect of the contact model on the impact-vibration response of continuous and discrete systems, J. Sound Vib. 332 (15) (2013)
3849–3878.
[7] M.R. Brake, The role of epistemic uncertainty of contact models in the design and optimization of mechanical systems with aleatoric uncertainty,
Nonlinear Dyn. 77 (3) (2014) 899–922.
[8] M.R. Brake, P. Reuss, D.J. Segalman, L. Gaul, Variability and repeatability of jointed structures with frictional interfaces. in: M. Allen, R.L. Mayes, D.J.
Rixen, (Eds.), Dynamics of Coupled Structures, vol. 1, Proceedings of the 32nd IMAC, A Conference and Exposition on Structural Dynamics, Springer,
2014, pp. 245–252.
[9] A. Caignot, P. Ladevèze, D. Néron, Prédiction par essais virtuels de l'amortissement dans les structures spatiales, in: Proceedings of the 8ème Colloque
National en Calcul des Structures, Giens, 2007.
[10] A. Caignot, P. Ladevèze, D. Néron, J.-F. Durand, Virtual testing for the prediction of damping in joints, Eng. Comput. 27 (5) (2010) 621–644.
[11] C. Cluzel, P. Rouch, O. Dorival, S. Behar, Experimental identification and associated model of the dissipation in the electric cables of a satellite structure,
J. Sound Vib. 357 (2015) 156–167.
[12] S. Daouk, F. Louf, O. Dorival, L. Champaney, On the lack-of-knowledge theory for low and high values of uncertainties. in: Proccedings of the 2nd
International Symposium on Uncertainty Quantification and Stochastic Modeling. Rouen, France, June 23–27, 2014.
[13] S. Daouk, F. Louf, O. Dorival, L. Champaney, S. Audebert, Uncertainties in structural dynamics: overview and comparative analysis of methods, Mech.
Ind. 16 (4) .
[14] J. De Crevoisier, N. Swiergiel, L. Champaney, F. Hild, Identification of in situ frictional properties of bolted assemblies with digital image correlation,
Exp. Mech. 52 (6) (2012) 561–572.
[15] D. Di Maio, Identification of dynamic nonlinearities of bolted structures using strain analysis. in: G. Kerschen, (Ed.), Nonlinear Dynamics, vol. 1,
Proceedings of the 33rd IMAC, A Conference and Exposition on Structural Dynamics, Springer, 2016, pp. 387–414.
[16] J. Esteban, C.A. Rogers, Energy dissipation through joints: theory and experiments, Comput. Struct. 75 (4) (2000) 347–359.
[17] D. Ewins, Modal Testing: Theory, Practice and Application, Research Studies Press Ltd, Baldock, England, 2000.
[18] L. Gaul, R. Nitsche, The role of friction in mechanical joints, Appl. Mech. Rev. 54 (2) (2001) 93–106.
[19] M. Géradin, D. Rixen, Mechanical Vibrations: Theory and Application to Structural Dynamics, Wiley, 2015.
[20] C. Hammami, E. Balmès, M. Guskov, Numerical design and test on an assembled structure of a bolted joint with viscoelastic damping, Mech. Syst. Signal
Process. 70 (2016) 714–724.
[21] M.W. Heinstein, D.J. Segalman, Bending effects in the frictional energy dissipation in lap joints. Tech. rep., Sandia National Labs, 2002.
[22] R. Ibrahim, C. Pettit, Uncertainties and dynamic problems of bolted joints and other fasteners, J. Sound Vib. 279 (3) (2005) 857–936.
[23] N.F. Knight Jr, D.R. Phillips, I.S. Raju, Simulating the structural response of a preloaded bolted joint. in: Proceedings of the 49th AIAA/ASME/ASCE/AHS/
ASC Structures, Structural Dynamics, and Materials Conference. 2008, pp. 1–20.
[24] S. Le Loch, Modélisation et identification de l'amortissement dans les structures spatiales (damping modelisation and identification in aerospace
structures), (Ph.D. thesis), ENS Cachan, France, 2003.
[25] M. Maheri, R. Adams, Modal vibration damping of anisotropic FRP laminates using Rayleigh-Ritz energy minimization scheme, J. Sound Vib. 259 (1)
(2003) 17–29.
[26] M. Maheri, R. Adams, J. Hugon, Vibration damping in sandwich panels, J. Mater. Sci. 43 (20) (2008) 6604–6618.
[27] B. O'Toole, K. Karpanan, M. Feghhi, Experimental and finite element analysis of preloaded bolted joints under impact loading. in: Proceedings of the
47th AIAA/ASME/ASCE/AHS/ASC Structures, Structural Dynamics, and Materials Conference, 2006, pp. 1–9.
[28] P. Rouch, C. Cluzel, O. Guezet, Meso-model and experimental identification of damping in joint of satellite sandwich panel. in: Proceedings of the
European Conference on Spacecraft Structures, Materials and Mechanical Testing, Toulouse, France, 2009.
[29] D.J. Segalman, Modelling joint friction in structural dynamics, Struct. Control Health Monit. 13 (1) (2006) 430–453.
[30] D.O. Smallwood, D. Gregory, R.G. Coleman, Damping investigations of a simplified frictional shear joint, in: Proceedings of the 71st Shock and Vi-
bration Symposium, SAVIAC, The Shock and Vibration Information Analysis Center, Citeseer, 2000.
[31] S. Smith, J. Bilbao-Ludena, S. Catalfamo, M. Brake, P. Reuß, C. Schwingshackl, The effects of boundary conditions, measurement techniques, and
excitation type on measurements of the properties of mechanical joints, in: G. Kerschen, (Ed.), Nonlinear Dynamics, vol. 1, Proceedings of the 33rd
IMAC, A Conference and Exposition on Structural Dynamics. Springer, 2016, pp. 415–431.
[32] E.E. Ungar, Energy dissipation at structural joints, mechanisms and magnitudes, Tech. rep., DTIC Document 1964.
[33] K. Van Buren, T. Hall, L. Gonzales, F. Hemez, S. Anton, A case study to quantify prediction bounds caused by model-form uncertainty of a portal frame,
Mech. Syst. Signal Process. 50–51 (2015) 11–26.
[34] H. Wang, Z. Qin, F. Chu, Effect of bolt number on joint stiffness of disc and drum connected by bolted joints, in: K. Chan, J. Yeh, (Eds.), Proceedings of
the International Conference on Electrical, Automation and Mechanical Engineering (EAME 2015), Atlantis Press, 2015, pp. 165–167.

ImportanteLeer EcuacionesEmpiricas

Uploaded by

Copyright:

Available Formats

You might also like

ImportanteLeer EcuacionesEmpiricas

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

ImportanteLeer EcuacionesEmpiricas

Uploaded by

Copyright:

Available Formats

Journal of Sound and Vibration ∎ (∎∎∎∎) ∎∎∎–∎∎∎

Contents lists available at ScienceDirect

Journal of Sound and Vibration

Study of the dynamic behavior of a bolted joint under heavy

1. Design of the assembly

1.1. Context of the study and related technical specifications

1.2. The bolted joint

1.3. Operating principle

use a beam type resonator to measure the loads in the joint.

1.4. Sizing procedure

2.1. Experimental evaluation of dissipation

2.1.1. Quality factor

In case of small damping, an approximation of the modal damping ratio ζ is:

2.1.2. Specific damping capacity (SDC)

Fig. 5. Illustration of the 3 dB bandwidth.

2.3. Experimental procedure

2.4. Experimental modal analysis of the assembly

77.7 Hz: first symmetric bending mode;

3. Results on the first bending mode

3.1. The various configurations tested

the reference configuration (i.e. with the reference part);

3.2. Evolution of ψ and the resonance frequency

Config. Low moment High moment Relative

8 bolts - 40 N.m 78.08 Hz 77.44 Hz 0.83 %

4. Exploitation of experimental results through differential analyses

4.1. Evolution of the apparent stiffness of the bolted joint

is minimized through a gradient descent with a variable step size:

Ei = argmin ⎡⎣ fi (E, k = k 0 ) − fi, mes ⎤⎦

The modulus E0 = 2.0 × 1011 N. m−2

ki = argmin ⎡⎣ fi (E = Ei, k ) − fi, mes ⎤⎦

4.2. Evolution of the dissipation in the bolted joint

5. Conclusions and future work

You might also like