Engineering Failure Analysis 141 (2022) 106581

Contents lists available at ScienceDirect

Engineering Failure Analysis

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

Stockbridge dampers for extending the fatigue life of a cable at St

Nazaire cable stayed bridge
Jacques Berthellemy a, *, 1, Dominique Siegert b, Édouard Berton c, Pierre Quentin d
a
Steel Bridges Engineering Expert at Cerema, France
b
COSYS/IMSE, University Gustave Eiffel, France
c
Engineer at Cerema, France
d
Engineer at Loire-Atlantique Local Authority, France

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

Keywords: The failure of a stay cable on the Saint-Nazaire bridge is investigated in detail. In 2000, fatigue
Stay cables cracks that had cut many wires led to a first replacement, alike to the initial, of the H32 stay
Wind cable. Examination of the deposited locked coil strand showed that fretting fatigue was involved
Fatigue
in the wire breaks. The cause of the failures at the anchorages is discussed/studied in depth in the
Vibration
paper. Bending moments of the rigid cable, at its exit from the anchorage socket were proven to
Damping
Stockbridge devices be the source of the fatigue stress due to vibrations induced by vortex detachments (VIV: Vortex
Induced Vibrations).
As wires ruptures occurred again in 2017 in the new H32 stay cable at the same place, further
studies on dampers were carried out. After numerical simulations, the “Stockbridge” damper
became a good candidate for attenuating the amplitude of vibrations. It could indeed be designed
to fit the required frequency range between 30 and 35 Hz.

1. Brief description of the cable stayed bridge

The Saint-Nazaire - Saint-Brévin bridge shown in Fig. 1 is a cable stayed bridge in service since 1975. Its total length is 3356 m and
it is made of:

• a central orthotropic steel deck with a length of 720 m;

• two prestressed concrete access viaducts, 1115 m long at the north and 1521 m long at the south.

72 stay cables are supporting the three orthotropic steel deck spans of 158 m, 404 m and 158 m. The upper cable anchors are of the
harp design type, and distributed on the upper segments of the towers.
The structure has a radius of curvature of 6428 m in its longitudinal profile, leading to a theoretical variation in altitude of
approximately 2,50 m between towers and center. The structure thus provides a minimum clearance height of 61 m above the lowest
water level in the 300 m wide channel area.
The steel box girder deck is made entirely of welded plates of constant thickness throughout the structure. It is stiffened by

* Corresponding author.
E-mail addresses: [email protected], [email protected] (J. Berthellemy).
1
Chairman of TC50 European Committee (Lighting Columns).

https://doi.org/10.1016/j.engfailanal.2022.106581
Received 5 January 2022; Received in revised form 27 May 2022; Accepted 23 June 2022
Available online 20 July 2022
1350-6307/© 2022 Published by Elsevier Ltd.
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 1. Aerial photograph of the entire structure.

Fig. 2a. Damage that occurred in 1998. Section of a wire cut by a fatigue crack.

transverse diaphragms every-four meters. Its width is 15 m and the interior height is of 3.2 m.
The thickness of the plates used varies between 12 and 16 mm according to the compression stresses in the deck.
The 12 m wide roadway includes 3 lanes of traffic: 2 lateral lanes of 3.15 m, a central lane of 3.20 m, and two lateral sidewalks.
The stay cables are designed with locked coiled ropes made of uncoated cold drawn steel wires. The outer diameters of the locked
coil stands are in the range of 72 mm to 105 mm with several layers of Z-shaped wires and their lengths are in the range from 72 to 169
m. The stay cables are subject to the prevailing westerly wind actions.

2. First failure of the cable H32 leading to its replacement in 2000

2.1. Description of the H32 cable failure

The 1998 failure of the H32 cable is shown in Fig. 2a. This cable H32 has a 72 mm diameter and a length of 126.6 m (distance from
hinge to hinge including the heavy 2.20 m long bottom socket carrying device). This is the first cable of the edge span that is not
anchored near the end pier but on the deck. The 4 cables above it have a much more strong section. It is therefore both one of the
longest cables and also the one with the smallest section.
The cable H32 has a steel section of 3568 mm2, a conventional modified steel density of 8.4 including the filling material and was
tensioned with a force of 1.55 MN. The tension measurement using the vibrating string model was obtained by using the second mode
with two waves to avoid any catenary effects. The structural damping ratio ξ of the cable was about 10-3. This quite high value for such
a long stay is related to the frictional dissipation near the hinge at the bottom anchorage that were not locked and still functioning. This

2
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 2b. Broken wire locations in the section of the strand. : Z shape wires, : cylindrical wires, : broken wires.

Table 1
Composition and distribution of broken wires in the locked coil strand.
Layer n◦ Wire shape Number of wires Number of broken wires

1 Z 51 38
2 Z 45 27
3 Z 38 26
4 O 30 17
5 O 24 7
6 O 18 1
7 O 12 0
8 O 6 0
Centre wire O 1 0

Fig. 3. Evidence of a fretting fatigue crack propagation cutting a Z wire section, while the brittle failure zone presents corrosion after several weeks
of exposure to the atmosphere.

cable was in service for about twenty-two years in 1998 when it appeared to be severely damaged with lots of wire breakages in the
external layer as shown in the Fig. 2b. The bridge was originally built without any damping device.
After the removal of the cable, the internal state of the stay cable at the sockets was investigated. The strand was made of 225 wires

3
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 4. Global monitoring of the bridge: ■ anemometers, • accelerometers, ★ traffic recorder.

Fig. 5a. Time record of the measured displacements for the vibrations of the replaced cable.

and 116 of them were found broken at the lower anchorage and only four at the upper one. The distribution of the wire ruptures in the
layers is given in Table 1.
All of the wire ruptures were located in a section between ten and thirty millimetres inside the threaded socket. Fig. 2b shows the
positions of the broken wires in the section and suggests the characteristic symmetry of the in plane bending condition of stress. The
neutral axis of the cross section for this bending is shown in Fig. 2b: it is in the direction of the prevailing westerly wind.
Observation of the fracture surfaces led to the conclusion that fatigue cracks were initiated at the inter-wire contacts or at the
contact between the external layer and the serving wire. Fig. 3 shows the surface fracture of a Z shaped wire. The surface fracture is
fresh enough to enable us to recognize the fretting fatigue crack propagation area initiated at the inter-wire contact.
The breaking of the wire on Fig. 3 occurs when the resisting surface has been sufficiently reduced by the fatigue crack propagation:
The remaining section on the right side is broken by brittle fracture, which provides to corrosion a very large surface area due to its
fractal geometry. A complete description of the failure was written by Brevet [1] after the investigation of the first wire ruptures
(1997–1998).

2.2. Monitoring of the bridge and vibration analysis

After visual inspection, the regular maintenance team had reported that the dismantled strand was vibrating in the plane
perpendicular to the prevailing westerly wind, at moderate amplitude but at quite high frequency. They had also noticed six months
earlier that the strand paint was damaged near the anchor.
These observations appeared to be vortex-induced vibrations (VIV). An extensive helpful literature review about wind induced
vibrations of stay cables had been written by Matsumoto [2] in 1995.
VIVs correspond to the alternating detachment of vortices on the upper and lower surfaces of the cross-sections, resulting in quasi-
periodic aerodynamic forces. When the frequency of this vortex shedding approaches a natural frequency of the structure, a lock-in
phenomenon occurs, which can cause the structure to oscillate with moderate amplitude. However, in the case of the H32 stay, the
natural frequency is high frequency above 10 Hz.
The bridge was therefore instrumented after replacement of the cable H32. The Fig. 4 shows the general instrumentation that we
put in place on the bridge.

4
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 5b. Related power spectral density of the records presented in Fig. 5b.

Fig. 6. Fretting fatigue experimental results for un-lubricated wires.

The monitoring set up shown in Fig. 4 consisted in measuring the movements of the main structural elements (deck, towers, stay
cables) in order to find correlations between the different structural movements and between traffic and wind actions. The monitoring
devices consisted of 25 accelerometers, three anemometers and an equipment for traffic analysis.
No correlation has been found between heavy traffic or tower and deck vibrations and the high amplitudes of vibrations of the
cables. However H32 is a west wind stay, not in very high tension because it is located just under four large retaining stays. In addition
H32 is on the west side of the bridge which is more exposed and prevailing winds lift H32 and reduce its tension.
The displacement of the new replaced H32 cable was recorded during four months in the vicinity of the lower anchorage during a
period of heavy vibrations at a distance de 685 mm along the cable. Fig. 5a shows an example of the displacement of the H32 cable
recorded during a period of cross wind excitation inducing VIV.
The corresponding power spectral density is shown in Fig. 5b.
The power spectral density of the recorded signal is displayed in Fig. 5b and indicates that the energy of vibration is mainly located
in the frequency range which is roughly between 30 Hz and 35 Hz.
In order to evaluate the bending deformations undergone by the stay cable in service conditions, a determination of the bending
strain based on the measure of the displacement has also been undertaken on the model of studies in great depth by Poffenberger and
Swart in 1965 [4]. According to this method of investigation, and as the result of the rainflow analysis, the replaced H32 cable was
mechanically affected, during the 4 months of records, by about 12,350 cycles and 2800 cycles with levels of stress of 80 MPa and 100
MPa respectively.
The fatigue limit of stay cables submitted to lateral vibrations has been previously addressed in the literature by Hobbs and

5
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 7a. Anti-vibration device proposed in 1999.

Fig. 7b. Cross-tie possible option.

Ghavami [3] and some collaborative research work on the fretting fatigue of steel wires was in progress at the École Centrale de Nantes
and Joseph Fourier University in Grenoble: Siegert (1997), Couroneau (1998), Delclos (2000), Olivella (2000).
Fretting fatigue tests on uncoated and non lubricated cold drawn eutectoid wires were carried out in the LCPC laboratory of Nantes.
These tests results displayed in Fig. 6 show that the value of the experimental fretting fatigue limit is about 100 MPa. The threshold is
therefore overcome in real service conditions, because the contact conditions at the socket of the cable constitute more aggressive
conditions than the laboratory test conditions.
The origin of wire breaks is clearly related to inter-wires fretting fatigue caused by wind-induced vibrations at frequencies around
30 Hz with bending of the wire at the point where the wire exits the adjustable threaded cylindrical socket. In addition, we have
noticed that some fretting fatigue initiations were located at the contacting surface between the cable and the socket.

2.3. Proposals to mitigate cable vibrations through damping solutions in 2000

Fatigue cracks occurred during 1998 have led to the first replacement of the cable H32 in 2000 alike to the initial cable. H32 new
cable is a full locked cable with an external diameter of made of helicoidal wires from ArcelorMittal Bourg-en-Bresse. However, there is
no final synthesis of all the studies carried out which would give a definitive sustainable solution for stay H32 after the 1998 failure and
no solution was found in 1999, that fulfilled the requirements of M. Virlogeux to strengthen the cable permanently.
The Fig. 7a presents the device designed at that time by J. C. Foucriat was rigid and we could prove that this device would have
increased the problem by preventing rotations at the foot of the anchor. The idea was however to block the rotations of the hinge that
always rotated perfectly at the foot of the heavy pivoting hub designed for the much more stronger cables of the bridge, and to use if
needed an intermediate damping device. J. Berthellemy also proposed a cross-tie solution to increase the insufficient tension of the
H32 cable while relieving the cables located immediately below, as shown in Fig. 7b. These more expensive options would be
considered if the dampers described in this paper are not sufficient to fix the vibration problem.
As new ruptures of some wires occurred in 2017, Cerema received after a call for tender, the mission to study the problem again.

3. First protection device studied to avoid cable flexion near the socket.

The first aim was to develop a device in order to preserve the structural integrity of the H32 cable stay. A global model of H32 was
built using PCP software (Cerema) and a local model using Code_Aster software (EdF-R&D: Code_Aster [5] is a general finite elements
Code for scientific calculations which is open source and certified for use in the nuclear industry in several countries). Meshes are

6
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 8. “Heavy” device more effective than the device presented in Fig. 7a to reduce the cable flexion where it comes out of the socket. The same
240 mm deformations of both figures are strongly amplified (representation from Code_Aster model).

Fig. 9. Half Stockbridge damper model.

realized with GMSH [6]. The PCP three-dimensional model uses beam elements, allowing consideration to the catenary effects and the
bending inertia of the cable. PCP means “Pont Construit par Phases” because this software is also able to take into account the sequence
of construction of the bridge. These phenomena must be included in the model to assess the effects of high frequency vibrations
induced by turbulent wind on the global behavior of the cable. More details regarding this model are given in section 5 of the present
paper.
Modal properties of the cable are evaluated, i.e. eigenfrequencies, mode shapes, modal masses and damping. The dynamic analyses
show that wind loads which frequency range from 10 Hz to 35 Hz induce maximum normal stresses at the cable ends.
The proposed device presented in Fig. 8 protects the location where the cable leaves the socket. However, as with the device
proposed by Foucriat in 1999, the new device simply moves the problem further along the cable. It is called “heavy device” in Fig. 8
because it is also heavy in comparison to the Stockbridge dampers that were designed to avoid cable flexion near the socket for the
frequency range between 10 and 35 Hz.

4. Preliminary design of the Stockbridge damper

In order to reduce the bending stresses in the stay-cable at the end fitting, the use of Stockbridge dampers is investigated for
increasing the damping in the frequency range of the vortex shedding excitation. The preliminary design parameters (geometry, mass
and stiffness distributions) of a Stockbridge damper were derived from a simplified model. In reference, this solution was studied and
implemented by Argentini et al. in [7] for the Hovenring bridge and very recently by Larsen et al. [8] for the Halogaland bridge.

4.1. Simplified model of a damper using a martyr messenger auxiliary cable (also called “Stockbridge device”)

The simplified 2DOF model of half of a symmetrical 2R type Stockbridge damper displayed in Fig. 9 is used for defining the design
parameters in the first design stage. The structure consists of a clamped beam OA which represents the messenger martyr cable and the
counterweight attached to it at the tip. The design parameters first considered with the simplifying assumption of a weightless
messenger martyr cable are the two resonant frequencies related to the bending deformation of the messenger cable in the plane xy
which is perpendicular to the cross-wind direction. These resonant frequencies depend on the length L and bending modulus EI of the
messenger cable and also on the mass m and rotational inertia J of the counterweight. The governing equations for this model were
derived by Claren and Diana [9], further theoretical and experimental investigations were carried out by Wagner et al. [10]. The

7
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

generalized displacement vector of the system is vT = [ vA θA ] where vA is the displacement in the y direction of the counterweight at
the point A and θA its rotation.
The equation of motion with the harmonic excitation FeiΩt and the proportional hysteretic damping βK for the damping model of
the messenger cable reads as:

(1)
Ấ
Mv + K(1 + iβ)v = FeiΩt
The mass matrix M and the stiffness matrix K are defined as.
⎡ ⎤
[ ] 4k 2k(L − lG )
m m(L − lG ) ⎢ ⎥
M= K=⎣ 4kL2 ⎦ (2)
m(L − lG ) J 2k(L − lG )
3

where k = 3EI
L3
is the stiffness of the cantilever beam OA. The rotational inertia J of the counterweight is defined at the point A and is
expressed using the radius of gyration rG of the counterweight and the distance between A and the center of inertia G of the coun­
terweight, so that lG represents the position of its center of mass.

J = mrG2 + m(L − lG )2 (3)

When the excitation is produced by the harmonic vertical displacement of the clamp VO e iΩt
the related amplitude of the force is
given by the row vector.
[ ]
FT = mΩ2 VO 0 (4)

Assuming a solution of the form v(t) = Veiωt for the free vibrations (FT = [ 0 0 ] of the undamped system leads to the eigenvalue
problem KX = ω2 MX for which the two solutions (ωr , rX) satisfy.
⃒ ⃒
⃒K − ω2 M ⃒ = 0 (5)

The forced vibration response v(t) = Ve iΩt

for the case of harmonic excitation reads.
( )− 1
V = K(1 + iβ) − Ω2 M F = α(Ω)F (6)

The matrix elements of the receptance frequency response function α(ω) can be expressed using the modal properties:
∑2 r Φj r Φk
∑2 Ajk
αjk = = r
(7)
r=1 ω2r − Ω2 − iβω2r r=1 ω2r − Ω2 − iβω2r

where the mode shape vectors are mass-normalised i.e.

ΦT MΦ = 1 (8)
At the resonance frequency there is equivalence between the hysteretic and viscous damping model taking β = 2ξ For the design of
an effective damping device the frequency interval [ω1 , ω2 ] must lie within the critical frequency range of Ω as for example in case of
vortex-induced vibrations of the cable attached to it.
The amplitude FO of the force acting on the clamped end at the point O in the y direction related to the harmonic vertical
displacement of the clamp VO eiΩt is given by.

FO = (1 + iβ)mVO Ω2 eT1 K α(Ω)e1 (9)

( )
1
where e1 = is a unit vector.
0
The impedance is defined as.
FO
Z(Ω) = = mΩ(β − i)eT1 K α(Ω)e1 (10)
iΩVO
Experimental estimates of Z(Ω) can be obtained in testing the Stockbridge damper with a shaker for different excitation amplitudes
in order to characterize the nonlinear behavior of the device. Further details about experimental determination of the impedance of
Stockbridge dampers can be found in [11] and [12].
It may also be used for predicting with a linear model the effectiveness of the Stockbridge damper to reduce the vibration am­
plitudes of the cable undergoing vortex-induced vibrations (see the following section about the numerical simulations with PCP).
According to the 2007 FHWA guidelines from Kumarasena et al., wind induced vibrations of the stay cable are prevented when its
structural damping is high enough over the excitation frequency range i.e. the values of the Scruton Sc number are above 10. The
Scruton number is defined by.

8
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Table 2
Values of the design parameters.
Counterweight mass m (kg) Counterweight J (kg.m2) CounterweightxG (mm) Martyr cable L (mm) Martyr cable EI (N.m2)

30.1 0.55 242.9 300 254

Table 3
Values of the modal constant r Ajk .

− 1 − 1 − 1 − 1
1 A11 (kg ) 2 A11 (kg ) 1 A12 (kg ) 2 A12 (kg )

0.024 0.019 − 0.08 0.3

Fig. 10. Modulus of the impedance.

ζn μ
Sc = (11)
ρD2

where ζn is the damping ratio related to the nth natural frequency of the stay cable, µ its mass per unit length, D its diameter and ρ the
air density. ζn is the addition of two components, one related to the cable, and one related to the additional damper: ζnd .
The damping ratio of the stay cable of length L between the fixed ends can be estimated as in reference [7,13] from the simplified
model of the Stockbridge damper located at the distance l from the fixed end. The contribution of the damper to the damping ratio is
given by the following formula:
Z(ωn ) ∨ cos(α(ωn ) )sin(nπl/LC )
ζnd = (12)
μLωn
This damping ratio is derived from the energy dissipated in the messenger cable of the Stockbridge damper during one cycle of
vibration and is added to the damping ratio of the stay cable without the damper device. The natural frequencies of the stay-cable are
given by the equation (13):
nπ √̅̅̅̅̅̅̅̅
ωn = T/μ (13)
LC

4.2. Numerical data of the preliminary design of the Stockbridge device

The data of the half Stockbridge model are reported in Table 2. They come from the proposal for the design of the damping device to
reduce the vortex induced vibrations of H32 stay cable on the bridge of Saint Nazaire. The messenger cable considered here is a two
layers spiral strand of 30 mm diameter made of 19 steel wires of 6 mm diameter. The bending modulus of the messenger cable is
calculated assuming the full slip regime between the helical wires in contact, leading to the lower bound estimate of the value for this
parameter.
The resonant frequencies calculated with the design parameters in Table 2 are 9.5 Hz and 13.9 Hz. These values are close enough
but are significantly below the target of frequency range related to the H32 VIV.
Instead of considering the full slip regime in the whole of the strand, when it is assumed that it takes place only at the outer layer of
the strand, the value of the bending modulus is then 875 N.m2 and the resonant frequencies are 17.8 Hz and 44.1 Hz. A better set of
design parameters is obtained for a messenger cable with a bending modulus of 500 N. m2 and a length of 200 mm. The resonant

9
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 11. Phase shift of the impedance.

Table 4
Parameters of the H32 stay-cable.
Length LC Lineic mass μ Tension T Diameter D l (m)
(m) (kg/m) (kN) (mm)

126.6 29.65 1550 72 1.65

Fig. 12. Scruton number of H32 stay-cable with a Stockbridge damper.

frequencies of half of the damping device are 24.3 Hz and 36.5 Hz which is more close to the expected range of critical frequencies of
VIV excitation than the previous results. The related values of the modal constant r Ajk of the Stockbridge model are reported in
Table 3.
The modulus and the phase of the impedance calculated for a loss factor of 0.16 are displayed in Figs. 10 and 11.
The values of the Scruton number are estimated in the case of the H32 stay-cable with the Stockbridge damper characterized by its
mechanical impedance Z. The parameters of the stay-cable are reported in Table 4 and the estimated values of the Scruton number are
displayed in Fig. 12.
The values calculated in the critical frequency range of the Karman wind induced vibrations are above 10 as expected for an
effective reduction of the vibration amplitudes when the Stockbridge damper is located at the distance l from the fixed end as already
defined. For low frequencies, the Scruton number remains low because of the low damping ratio of the cable, and because the
Stockbridge damper is not effective in the low frequency range.

5. Preliminary analysis using beam modeling

The modelling of the H32 stay has three main objectives: (i) evaluate the influence of stays’ flexural rigidities, (ii) evaluate the
influence of higher frequency modes and (iii) evaluate the efficiency of a damping system, such as Stockbridge, to reduce vibrations in
the H32 stay. H32 stay is modelled alone with its sockets, as previous studies have shown that deck vibrations on H32 stay dynamic

10
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 13. Modal shapes of the first 4 modes: (a) first vertical mode (0.86 Hz), (b) first transversal mode (0.88 Hz), (c) second vertical mode (1.70 Hz)
and (d) second transversal mode (1.77 Hz).

behavior are negligible.

The first model considers the catenary effects so that the deformed shape of the stay cable under its own weight is considered, but
the flexural rigidity is neglected. The model is composed of two rigid elements modelling the deck and the mast, and two hinges
allowing the rotation around a horizontal axis, between the lower socket and the deck, and between the upper socket and the mast. The
socket has the following mechanical properties, obtained through a finite element modelling: a cross-section of 7.744.10-2 m2, a
torsional inertia of 9.544.10-4 m4 and a flexural inertia of 4.772.10-4 m4. The stay cable has a cross-section of 3.57.10-3 m2, a mass per
unit length of 29.4 kg/m and is discretized by 10 cm long element over three meters at both extremities and by 1-meter-long element
over the rest of its length. However, according to Virlogeux’s report 2000, higher frequency modes put the socket and the stay cable out
of phase. Because of its inertia, the socket does not vibrate under high frequency solicitations, causing an angular break with the stay
cable. A second approach is achieved, using beam elements to model the stay cable to consider flexural rigidity. The first model,
considering the catenary effect, is used to calibrate the beam model regarding its deformed shape and tension in the stay cable. Beam
elements modelling the stay cable have a cross-section of 3.568.10-3 m2, a torsional inertia of 2.64.10-6◦ m4, a flexural inertia of
1.32.10-6 m4 and a density of 8.24 t/m3. All other elements are identical to the first model. Tensioning of the beam stay-cable is
performed through a fictional thermal shrinkage of − 247.517 ◦ C, calibrated by a convergence algorithm to obtain a tension and a
deformed shape identical to the first model with a precision of 10-1 Pa. The first 100 modes are evaluated for both models: gaps range
from 0 % to 5.8 %. It is observed that the flexural inertia tends to increase frequency of modes. We observe that the frequency of the
first mode, identical for both models (0,86 Hz), is very close (0.88 Hz) to the one obtained experimentally by LCPC and Advitam, with a
difference of 2.3 %. Fig. 13 shows the modal shape of the four first modes.
Two previous studies by Virlogeux (2000) and Brevet (2002) showed that tendons rupture in the vertical plan can be attributed to
high frequency vibrations induced by vortex shedding, between 28 Hz and 40 Hz. The study of this phenomenon is performed based on
prescriptions of Annex E of Eurocode 1–1-4 [14]. The resonance occurs when the frequency of the vortex shedding equals the fre­
quency of a mode of the stay cable. The frequency f of the vortex shedding is given by:
U
f = St (14)
D

where U is the average speed of the wind far from the stay cable, D is the diameter of the stay cable (0.072 m) and St is the Strouhal
number, which value is 0,18 according to Annex E of Eurocode 1-1-4 [14]. A temporal analysis is conducted with forces induced by
vortex shedding applied on points of the stay cable where modal displacement are maximum, for two significant modes: mode 67, with
a frequency of 29.50 Hz, and mode 78, with a frequency of 35.06 Hz. The frequencies of these two modes correspond to the bounds of
the frequency range identified in previous studies by Virlogeux (2000) and Brevet (2002) to cause high frequency vibrations in the
vertical plane of H32. The value of the force applied on each node with maximum modal displacement is determined with:

ρDU 2
F= CL sin(2πnt) (15)
2

where ρ is the density of air (ρ = 1.35 kg/m3) and CL is the lift coefficient, which value is 0.6 if Reynold number is between 40 and
3.0.105, and t is the time. Reynold number is given by:
UD
Re = ρ (16)
μ

11
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 14a. Bending moment at the junction between the bottom socket and the stay-cable for mode 67.

Fig. 14b. Bending moment at the junction between the bottom socket and the stay-cable for mode 78.

Fig. 15. Simplified modelling of the Stockbridge damper, adapted from [15].

where μ is the dynamic viscosity of air. For modes 67 and 78 of H32 stay-cable, Reynold number is between 1650 and 90 000, con­
firming the value of 0.6 for the lift coefficient CL .
Effects of high frequency vibrations induced by vortex shedding are evaluated at the junction between the bottom socket and the
stay-cable, in terms of stress variation caused by bending moment:
ΔM
Δσ = v (17)
I

where v = D/2 and I = π D4 /64, corresponding to the flexural inertia of the stay-cable.
ΔM is calculated from M values provided by the finite element PCP computation.
The inertia I of the cable was calculated using the hypothesis of inter-wires no-slippage. Figs. 14a and 14b show the variation of
bending moment over time at the junction between the bottom socket and the stay-cable for vortex shedding forces corresponding to
modes 67 and 78. For mode 67, the maximum bending moment variation of 2 kN.m is equivalent to a stress amplitude of about 100
MPa. For mode 78, the maximum bending moment variation is equivalent to a stress amplitude of about 90 MPa. These significant

12
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 16. Modes 1 to 6 of the martyr cable Stockbridge prototype. Frequency map obtained by the Code_Aster model:

Fig. 17. Modes 7 to 10 of the martyr cable Stockbridge prototype.

stress variations at the junction between the socket and the stay-cable can explain the tendon break witnessed on H32.
To moderate stress amplitudes induced by vortex shedding, this study proposes to design and study a Stockbridge damper, solution
widely used to reduce vibrations on high voltage power lines. One initial approach consists in a simplified modelling of the Stockbridge
damper [15] presented in Fig. 15 where Mc is the mass of the stay cable, k is the stiffness of the Stockbridge damper, cd is the damping
coefficient of the Stockbridge damper and m is the mass of the Stockbridge damper. Here Z is the local reference axis perpendicular to
the cable.
The Stockbridge damper is modelled in PCP Software with a hinge element associated with a specific stiffness matrix and an
average damping coefficient. Adequate dynamic properties of the Stockbridge damper will have to be determined with an experi­
mental program. The temporal analysis conducted earlier to assess the effects of high frequency vibrations induced by vortex shedding
is repeated to assess the efficiency of the damper solution. Based on the same methodology, stress amplitude at the junction between
the bottom socket and the stay-cable for vortex shedding forces corresponding to mode 67 is reduced to 26.2 MPa, representing a

13
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 18. Distribution of the natural frequencies of the Stockbridge martyr cable damper.

Fig. 19. Cross section of the martyr rope.

decrease of more than 73 %. For vortex shedding forces corresponding to mode 78, stress amplitude is reduced by almost 90 % (9.8
MPa). This preliminary analysis show the great potential of a Stockbridge damper to reduce high frequency vibrations induced by
vortex shedding.

6. Design of the martyr cable Stockbridge damper

After the previous studies whose validity is most often based on linear models, the mechanical characteristics of the dampers have
to be determined and then tested in situ. Complementary elements can be found in the literature in Zasso et al. [7,13], Shan Huang [16]
and Hagedorn et al. [12]. Final fine-tuning of a damper should be done in situ, for example by adjusting the distances between the two
arms.

6.1. Determining geometry and characteristics of the damper

However, it must be precisely calculated in order to target the right range of frequencies where the vibrations to be avoided occur.
That’s why numerical simulations with a more detailed finite element Code_Aster model were carried out to get more precisely the
locations of the relevant resonant frequencies of the Stockbridge damper.
Figs. 16 and 17 show the calculated vibration eigenmodes for a Stockbridge device.
In Fig. 18, the martyr cable Stockbridge damper is designed so that its natural frequencies are distributed as evenly as possible in the
10–35 Hz range.
Any necessary adjustment of the frequency range above 35 Hz is easy by slightly reducing the length of the martyr cable.
After the preliminary design phase, the dynamic characteristics of the Stockbridge damper (free response spectrum and impedance)
will be verified in the laboratory and then in situ with calibrated and certified measuring equipment. The distance between Stockbridge
device and anchor is related to the position of the “antinodes” of the modes of vibration of the stay cable e.g. the points that move with
the greatest amplitude (see Fig. 21) that are critical regarding the damaging vibrations and will be also verified in situ. This issue is
described in more detail in the section 6.3.

6.2. Component parts of a damper unit

The components of the Stockbridge shock absorber are as follows (see Fig. 20):

14
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 20. Schema and details of a damper unit.

Fig. 21. Modal deformation of H32 for its mode n◦ 33.

Fig. 22. Complete system of 3 Stockbridge damping unit with martyr cables.

• Tie: A bite adapted to the diameter of H32 (identical to those used when the rope was first changed).
• Martyr rope: A multi-layered strand consisting of a central core wire plus at least two layers of helical wires on a layers of helical
wires over a length of 1 m (e.g. central part of a rope similar to H32 itself for example).

15
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

Fig. 23. General layout of the three martyrs’ cable devices arranged at 90◦ and in such a way as to reduce drag.

Martyr rope cross section is on Fig. 19. Only spiral strand cables should be used. Lifting and lift cables are too flexible and should be
avoided.

• 4 Jaws: Adapted to the diameter of the martyr rope

(3 + 1 guard jaw device at the bottom that also introduces dissymmetry

• Arms: Very rigid and more or less massive arms (3 sets)

• Rings: Two rings each carrying a mass of ~ 8 kg (3 sets)
• Collars: Collars according to the needs to prevent the wires from sliding

by adjusting the friction between the wires.

6.3. Number and disposition of the martyr cable dampers

The position of the Stockbridge devices is defined primarily by the desire to avoid the nodes of the mode shapes of the H32 cable.
The position was refined by using the PCP results (see Fig. 21).
The average distance from the first belly of the modes is 3.68 m. As an example, the modal deformation of H32 for its mode 33
calculated by PCP is shown.
The number of Stockbridge units is a minimum of three units rotated at 90◦ so that all modes of the Stockbridge units are activated
by the movements of the wind-driven H32 cable in each of the two possible basic directions. Two units are coupled and positioned
laterally to balance each other (Fig. 22). This prevents them from rotating as a result of H32′ s agitation.
The average distance between the socket and the group remains of 3.68 m. The distance between the attachment points on H32 is
determined to 1300 mm to avoid any contact.
The devices at the top can be rotated slightly to increase the number of modes involved: an angle of 20◦ has been shown for
illustration purposes on Fig. 23. These details of the layout are to be determined during the adjustments, by checking the aerodynamic
behavior of the devices attached to the shroud.
Recent publications [17] by Sébastien Langlois (University of Sherbrooke) and Frédéric Legeron, who monitored the imple­
mentation of Stockbridge dampers on large cables for Hydro-Quebec, confirms that only experimental tests can accurately assess the
dynamic properties of a Stockbridge prototype and carry out the final adjustments in situ.

7. Conclusion

Previous investigations into the failure of the H32 stay cable presented in this paper have shown that this rope is sensitive to vortex-
induced vibrations (VIV) that lead to wire breaks due to fretting fatigue. In the frequency range between 30 Hz and 35 Hz where these
VIVs occur, its own structural damping is too low. Numerical models of the H32 stay cable with the dampers validated their beneficial

16
J. Berthellemy et al. Engineering Failure Analysis 141 (2022) 106581

effect on the dynamic behavior under the VIVs excitation conditions.

Without “Stockbridge” damping devices deployment using messenger cables, the H32 cable would become weaker with each new
fatigue crack affecting a wire because of the vortex shedding vibrations. It could have to be changed again in the event of a very
aggressive storm similar to the one that occurred in 1998. This would not be a sustainable solution. However, it should finally be noted
that even in this case, all the damping devices provided in this article could eventually be mounted on a third new H32 stay cable after
its change, so as to prevent this VIVs problem.
Messenger cables may be called martyr cables because they endure fatigue damage in place of the main cable. However it is not
possible to accurately estimate the life of each martyr device in advance. In next future, the implementation of “Stockbridge” damping
devices using martyr cables will enable to verify in situ their efficiency to increase significantly the fatigue life of the stay H32 of the
Saint-Nazaire bridge by reducing its vibration amplitudes in the frequency range between 30 Hz and 35 Hz, where its own damping is
not sufficient. After installation on the cable with the necessary fine-tuning of the devices, monitoring will evaluate their effectiveness,
and future inspections will verify their effects on the fatigue life of the cable.

Declaration of Competing Interest

The authors declare that they have no known competing financial interests or personal relationships that could have appeared to
influence the work reported in this paper.

Acknowledgements

Acknowledgements to the bridge owner, Conseil Départemental de Loire-Atlantique, who accepted the publication of details for
scientific evaluation and knowledge diffusion.

References

[1] D. Siegert, P. Brevet, J. Royer, Failure mechanisms in spiral strands under cyclic flexural loading close to termination, OIPEEC Round Table, (Reading
Sept 1997): 111–119.
[2] M. Matsumoto, H. Ishizaki, J. Kitazawa, J. Aoki, D. Fujii, Cable aerodynamics and its stabilization, Int. Symposium On Cable Dynamics, Liege, Belgium, 1995.
[3] R.E. Hobbs, K. Ghavani, The fatigue of structural wire strands, Int. J. Fatigue 4 (2) (1982) 69–72.
[4] J.C. Poffenberger, R.L. Swart, Differential displacement and dynamic conductor strain, IEEE Trans., PAS-84, 281-289 (1965).
[5] de France, Électricité. “Finite element code aster, analysis of structures and thermomechanics for studies and research, (1989–2017).” Open source on www.
code-aster. org.
[6] Ch. Geuzaine, J.-F. Remacle, GMSH: A finite element mesh generator with built-in pre- and post-processing facilities, (2017).
[7] T. Argentini, L. Rosa, A. Zasso, Experimental evaluation of Hovenring bridge stay-cable vibration, WIT transactions on modelling and simulation 55, 427–437.
[8] H. Larsen, K.G. Andersen, A. Jamal, Wind induced hanger vibrations, the Halogaland suspension bridge, Structural Engineering International Nr. (1/2022.).
[9] R. Claren, G. Diana, Mathematical analysis of transmission line vibration, IEEE Trans. Power Apparatus Syst. (12) (1969) 1741–1771.
[10] H. Wagner, V. Ramamurti, R.V.R. Sastry, K. Hartmann, Dynamics of Stockbridge dampers, J. Sound Vibration 30 (2) (1973) 207-IN2.
[11] P. Hagedorn, On the computation of damped wind-excited vibrations of overhead transmission lines, J. Sound Vib. 83 (2) (1982) 253–271.
[12] P. Hagedorn, N. Mitra, T. Hadulla, vortex–excited vibrations in bundled conductors: a mathematical model, J. Fluids Struct. 16 (7) (2002) 843–854.
[13] T. Argentini, L. Rosa, A. Zasso, Wind-induced vibrations of the stay cables of a roundabout flyover: Assessment, repair, and countermeasures, J. Perform. Constr.
Facil 30 (6) (2016) 04016046.
[14] NF EN 1991-1-4, “Actions on structures — Part 1-4: General actions — Wind actions”.
[15] O.R. Barry, et al., Vibration Modeling and Analysis of a Single Conductor with Stockbridge Dampers, University of Toronto ProQuest Dissertations Publishing,
Degree (2014). 3630214.
[16] S. Huang, Vortex Induced Vibration (VIV) suppression of a two-degree-of-freedom circular cylinder and drag reduction of a fixed circular cylinder by the use of
helical grooves, J. Fluids Struct. 27 (7) (2011) 1124–1133.
[17] S. Langlois, F. Légeron, Prediction of Aeolian Vibration on Transmission-Line Conductors Using a Nonlinear Time History Model - Part I: Damper Model and Part
II: Conductor and Damper Model, IEEE Trans. Power Delivery 29 (3) (June 2014).

17