Professional Documents
Culture Documents
02 11 ULB Bouillard PDF
02 11 ULB Bouillard PDF
Summary
The study of the effects caused by human excitation on structures, mainly vibrations,
has gained a significant evolution during the last few years. Some shortcomings have
been highlighted in existing codes of practice in relation to the dynamic response of
footbridges but nevertheless, there is still no clear regulatory guidance on dynamic
design of pedestrian bridges.
With current design practice for footbridges, vibrations are becoming an important
issue. This is due to several reasons such as high resistant materials, smaller cross
sections or larger spans. All this causes a reduction of the stiffness leading to smaller
natural frequencies and therefore the structure exhibits a higher risk of resonance with
pedestrian excitation. As a consequence, the vibration issue becomes now a main reason
for extending the design process to dynamic loads.
In order to take the dynamic behaviour into account at the stage of the design study, it is
necessary to model pedestrian loading on the footbridge, which result from rhythmical
body motions of persons. Vibrations of footbridges lead mainly to serviceability issues
such as the comfort of the pedestrians. In order to verify that our structure does not
reach such a serviceability limit state, the verification criteria are given at the end of this
article highlighting the most appropriate ones for the College footbridge.
When considering the vibration behaviour under pedestrian loading, the major
parameter is the stepping rate (frequency) of pedestrians. Typical pacing frequencies for
a person walking generally lie between 1,6 and 2,4 Hz with a mean value of 2 Hz. 50%
of pedestrians walk at rates between 1,9 and 2,1 Hz and 95% of pedestrians walk at
rates between 1,65 and 2,35 Hz.
Rhythmical human motions during at least 20s tend to lead to almost periodic dynamic
forces, which can be described by a Fourier series of the following form:
n
Fp (t ) = G + ∑Gα i sin (2π i f p - φ i ) (1)
i=1
where G is the average weight of a person, fp is the pacing frequency and φi is the phase
B B B B
lag of the i-th harmonic. Different authors will give or not a special form to this general
Fourier series, each one adapting the different intervening parameters to their studies or
measurements. See [1], [2], [7] or [10].
Figure 2 represents the time function corresponding to the dynamic load created by a
person walking at a rate of 2 Hz. Not only the frequency of the first harmonic of the
forcing function (which is equal to the pacing rate) is of relevant importance, but also
the frequencies of upper harmonics.
Figure 2 : Time function of the vertical force from walking at a pacing frequency of 2 Hz [1]
Footbridges are commonly excited by several persons. Two different situations may
occur:
random action: the pacing frequency of the pedestrians is distributed according to a
U U
probability curve, while the phase angle of the 1st harmonic is characterized by a
P P
phase. This results in an increase of the force which in the case of the 1st harmonic is P P
It is very difficult to evaluate at each moment which type of situation is taking place. In
general, scientists agree that for small number of persons (not exceeding 15 to 25
individuals), a relatively perfect synchronization can be assumed.
The lock-in effect describes the phenomenon by which, when the structure exceeds a
certain threshold value of displacement (depending on direction of vibration, type of
activity, etc.), a walking (or jumping, or running) person tends to adapt to and
synchronize his/her motions in frequency and phase with the vibrating deck. If the
individual limit value of displacement is exceeded, then the user tends to give a certain
impulse into every vibration wave of the bridge. As a consequence, the vibration
amplitude of the structure increases and due to this, more and more persons are
“locked” into synchronization.
Values of this threshold have been proposed by several authors and will be exposed
further below in section 7. The order of magnitude stands around 10 to 20mm for
vibration amplitude. Practice codes contain however very little or none information
regarding this phenomenon.
It is possible to find in the literature several dynamic models representing the person’s
rhythmical body motion. Among them, Kreuzinger [9] and Stoyanoff [11][15] remain
sensitively close concerning the characteristic of a single pedestrian, but differ when
describing the effects caused by crowd loading. Bachmann’s [1][2] and Petersen’s
studies [10] seem nowadays to be the most complete and convincing among all others.
The main difference between both of them relays on the appraisal of the normalised
dynamic force, which depends significantly on the pacing frequency in the case of
Petersen.
The choice of the dynamic model was made taking account of factors such as validity of
the proposed model, level of detail, correct and feasible appraisal of all intervening
parameters and easy implementation info a finite element procedure. The decision
favoured Bachmann’s theory completed by Petersen’s model.
Petersen gives the following formulae for describing the forcing function created by a
single pedestrian:
Fp (t ) = G + c 1 G sin (2π f p t ) + c 2 G sin (4π f p t − ϕ 2 ) + c 3 G sin (6π f p t − ϕ 3 ) (2)
On the contrary to Bachmann, Petersen gives values of the different parameters which
vary significantly with respect to pacing frequency. Table 1 shows the Fourier
coefficients and phase lags proposed by Petersen for the case of a single pedestrian
excitation for 3 different values of excitation frequency.
Table 1
Petersen’s parameters and coefficients [10]
G weight of pedestrian N
fp
B pacing frequency
B Hz
fp = 1.5Hz
B B fp = 2.0Hz
B B fp = 2.5Hz
B B
According to references [1], [2] and [10], for a footbridge with a simple beam
behaviour, assumed to be supported by two simple bearings, whose natural frequency
remains in the range 1.8 to 2.2Hz, and for a random pedestrian action, the vibration
amplitude of the excitation force caused by a single pedestrian can be multiplied by an
enhancement coefficient m:
m = λT0 (3)
where λ is the mean flow rate of the pedestrians for a certain period of time and T0 is the
B B
In the case where the structure first natural frequency is not close to the mean value
2Hz, or when the synchronization between pedestrians cannot be considered as perfect,
the m factor has to be reduced [1] [10].
Bachmann’s magnification factor m is only valid for a bridge with assumed simple
beam behaviour and supported only by two simple bearings. For the College footbridge,
this is not the case. Indeed, the structure has an S-form and presents therefore an
important torsional component in its behaviour. In addition to this, the footbridge is
supported by two simple bearings at both ends, but also by two elastic bearings at 100m
approximately and respectively from extremities (suspended cable).
A complete analysis and the description of the method used to obtain the reduction
formula is exposed in [16]. The reduction formula, as a function of the ratio d/L (figure
3), obtained for different topologies of hyperstatic bridges is the following:
2
⎛d ⎞ ⎛d ⎞ (4)
reduction = − 7,8483π ⎜ ⎟ + 3,8255 ⎜ ⎟ + 0, 0114
L
⎝ ⎠ ⎝L⎠
The final aim of this work is to perform a dynamic analysis with the help of SAMCEF
(MECANO) in order to evaluate whether the College footbridge satisfies or not
verification criteria concerning vibrations for pedestrian excitation. The forcing function
(Fourier decomposition) will be applied at mid-span (where displacements and
accelerations are maximum) magnified by the enhancement coefficient m given by
Bachmann’s generalized theory.
The damping characteristics of the structure are of non negligible importance within the
dynamic analysis. The parameter that describes at best the dissipating behaviour of a
structure is the critical damping coefficient, which will be developed below.
The finite element procedure used in the computations is SAMCEF. In order to carry
out the natural frequencies computation, with module DYNAM, it is necessary to
initially perform a non-linear structural computation. This non-linear computation has
for goal to introduce the self-weight stresses into the structure cables and therefore
increase their stiffness. As a consequence the natural frequencies slightly increase. It
was not possible to perform the dynamic analysis with REPDYN, which would have
helped us to carry out a modal analysis, due to the significant non-linearities introduced
by the suspension cables in the structure.
Geometry
The main structure of the footbridge is modelled by means of shell, beam and rod
elements, while the approach span is modelled only with beam elements. The cross-
section is actually composed of 5 to 10mm thickness steel plates, which are stiffened in
order to absorb local loads. Concerning the global mechanic behaviour of the bridge, the
analysis is only concerned by the bending (and torsional) stiffness of the cross-section,
and not its local behaviour. This is why the stiffeners are not included in the finite
element model. However, this could cause an instability (warping) of the plates, and as
a solution, the plate thickness has been artificially increased by a multiplication factor
equal to 10, and the elasticity modulus has been decreased by 10 as well. As a
consequence, the beam and torsional stiffness of the plates remains unchanged.
Concerning the main suspension cables and the lower line of the deck, these have been
modelled using beam elements.
Loading
The only loads applied to the footbridge are the weight load, which is already
introduced into the model for the natural frequencies computation, and the pedestrian
load given by Petersen’s formulae (see section 6.5). For a correct appraisal of the self-
weight, the density of all the materials defined in the finite element model has been
modified in order to automatically introduce the load coming from the deck concrete
coating, the railings and the stiffeners which do not appear as independent elements in
the model.
The mesh is composed of a total of 16590 nodes, which gives a model characterized by
49770 degrees of freedom. The total number of elements contained in the model is
2446, of which 9 are rod elements, 1776 are 4-node shell elements, 36 are 3-node shell
elements and finally 625 are beam elements.
The value of the damping coefficient c affects the decay of the resulting vibration but
has weak influence on its frequency. If c > 0, the non-dimensional damping ratio is
defined as:
c c
ζ = = π (fn = natural frequency) (5)
c crit mf n
B B
By carrying out the analysis with SAMCEF, it is not possible to introduce structural
damping by means of the critical damping ratio ( ζ ). This is normally possible with the
option .AMO but version 9.1 does not allow the use of .AMO unless the option NLIS -1
is included in the data file. The option NLIS suppresses from the analysis all
geometrical non-linearities, which is not compatible with our structure due to the
suspension cables. For this reason, the only way to take into account natural damping is
to define a dissipating visco-elastic behaviour for the materials used in the finite
element modelling.
The material law of a visco-elastic material depends on the strain velocity. There are
several types of visco-elastic models. The model chosen for the current study is
described by Kelvin [12]. The equations relating stresses to strains are the following:
The visco-elastic coefficient α characterises the behaviour law from the material and it
is the only way in which it is possible to introduce damping into the structure.
Currently, there are very few references in public or academic literature concerning the
value of α and therefore it appeared necessary to establish a certain correspondence
between this coefficient and the critical damping ratio ( ζ ) which is the typical
parameter used in references.
c c 2 km
m&x& + kαx& + kx = 0 ⇔ c = kα ⇔ α= = ζ crit = ζ (7)
k k k
Finally:
ζ = π α fn (8)
where fn is the first natural frequency of the structure.
B B
Nevertheless, if we consider that the analysis is carried out in the modal basis, and that
material damping is proportional to mass and stiffness ( c = kα + m β ), then the equation
for each of the natural modes would be the following [13]:
The critical damping ratio ζ for a certain mode i can be calculated in the following
way:
β αωi
ζi = + (10)
2ωi 2
In the case we are concerned by, the damping is only proportional to stiffness, which
means that β is equal to zero. In that case, the critical damping ratio is a linear function
of ω , showed in figure 5. Thus, the material damping is a function of the excitation
frequency, and therefore it is impossible to fix the visco-elastic coefficient α to a single
value for all frequencies.
ζ
α
2 ω
Figure 5: Relation between the critical damping ratio and the eigen pulsation
The results will be presented as a table showing values of displacement and acceleration
due to the periodic pedestrian function applied at mid-span (and magnified by m), for 5
different points in the structure. The distribution of these “sensitive” points is given in
figure 6 below.
Point P1 Point P2
Nœud 15115 Nœud 5388
Point 73
Point 44 Nœud 7324
Nœud 2663
Point 15
Nœud 13746
Point LH0
Nœud 15269
These points of the structure are considered to be “sensitive” because they present the
higher displacement amplitudes or accelerations among all other significant points of
the structure. Points P1 and P2 were kept in order to verify that the supporting cables do
not undergo severe movements.
Using the reduction formula (4) obtained in [16], it is possible to calculate several
values of the amplification coefficient m. Petersen [10] suggests different values for
pedestrian density and velocity with respect to frequency, which were used to calculate
the number of persons (n) standing on the bridge for different situations. A final value
of 13 was chosen for the enhancement factor m.
6.3 Damping
Reference mean values for the critical damping ratio stand around 0.005 for steel and
0.010 for concrete, so that for the College footbridge (whose 22% of the total mass is
made of concrete) the appropriate value equals 0.006. Indeed, the main structure of the
footbridge is made of steel but the deck supports a 12cm thickness concrete slab along
approximately 86m.
A dynamic modal analysis was carried out in order to calculate the natural frequencies
from the College footbridge. This analysis is of crucial importance to predict which
pacing frequencies will cause high displacement or acceleration amplitudes. The
analysis was carried out in SAMCEF with DYNAM. The results are contained in table 6
below.
Table 5
College Brug natural frequencies
Mode Frequency (Hz)
1 0.79489
2 0.93805
3 1.21240
4 1.53988
5 1.75637
6 1.82897
7 2.21344
8 2.55394
9 2.67660
10 2.82364
11 2.98122
12 3.05543
13 3.06589
14 3.19878
15 3.28652
The College footbridge natural modes are complicate to identify. There are no pure
bending nor torsional modes and two different types of deformation always appear at
the same time, principally due to the length and relative orientation of the lateral spans
in relation with the central span. Figure 7 shows the undeformed shape as well as the
deformed shapes corresponding to modes 1 up to 5.
UNDEFORMED SHAPE
MODE
MODE 2
MODE 3
MODE 4
MODE 5
We decided to carry out dynamic computations for the following the frequency range
1.68 to 2.3Hz, which allows us to consider at least 97% of all probable pacing
frequencies. In addition to this, we noticed by analysing the natural frequencies. Table 6
contains the Fourier coefficients (according to Petersen’s formulae) which were used in
the computations.
Table 6
Petersen’s Fourier coefficients for frequencies 1.68 to 2.3Hz
fexc (Hz)
B B c1
B B c2
B B c3
B B φ1 = φ2
1.68 0.194 0.117 0.018 0.226
1.70 0.207 0.114 0.018 0.251
1.71 0.214 0.113 0.018 0.264
1.72 0.220 0.112 0.018 0.276
1.75 0.241 0.109 0.018 0.314
1.78 0.261 0.105 0.018 0.352
1.80 0.274 0.103 0.018 0.377
1.84 0.301 0.098 0.018 0.427
1.90 0.341 0.091 0.018 0.503
1.94 0.386 0.086 0.018 0.553
2.00 0.408 0.079 0.018 0.628
2.30 0.474 0.066 0.032 1.005
Tables 7 and 8 and figures 8 and 9 contain the results (nodal vertical displacement and
acceleration) obtained with the finite element software used for the dynamic analysis
(SAMCEF) for the range of excitation frequencies in between 1.68 and 2.3Hz.
Table 7
Computational results. Nodal displacement for frequencies in the range 1.68 to 2.3Hz
vertical displacement (mm)
fexc (Hz)
B B
node 2663 node 15115 node 5388 node 7324 node 15199
1.68 6.600 0.298 0.626 6.670 7.700
1.70 7.050 0.301 0.675 7.220 8.200
1.71 12.400 0.570 1.200 14.000 14.050
1.72 12.300 0.525 1.215 14.450 13.450
1.75 6.480 0.331 0.667 8.550 6.780
1.78 5.040 0.252 0.440 5.410 5.590
1.80 4.240 0.212 0.371 4.030 5.100
1.84 3.180 0.154 0.277 3.100 3.700
1.90 2.530 0.118 0.210 2.570 3.190
1.94 2.485 0.111 0.221 2.280 3.080
2.00 2.290 0.172 0.270 2.600 2.880
2.30 1.650 0.175 0.335 1.355 1.545
Table 8
Computational results. Nodal acceleration for frequencies in the range 1.68 to 2.3Hz
vertical acceleration (m/s²)
fexc (Hz)
B B
node 2663 node 15115 node 5388 node 7324 node 15199
1.68 0.395 0.083 0.098 0.400 0.560
1.70 0.425 0.086 0.097 0.420 0.575
1.71 0.780 0.108 0.099 0.850 0.870
1.72 0.760 0.108 0.100 0.890 0.930
The nodes presenting the largest displacement amplitude and acceleration are nodes
2663, 7324 and 15199. Indeed, these nodes are situated at mid-central-span and at mid-
lateral-span of the footbridge, respectively, where the structure is the most slender.
Displacements for nodes 15115 and 5388, situated at the base of the two cable supports,
remain negligible, which is reassuring given that both cables are supposed not to absorb
any important motions.
The results show a very important peak for both displacement and acceleration for an
excitation frequency of 1,72Hz. This is due to resonance with the fifth natural mode of
the structure and to damping (causing the natural frequencies to slightly decrease).
16,0
7324
14,0
5388
15115
2663
10,0
node 2663
8,0 node 15115
node 5388
node 7324
6,0 node 15199
4,0 - ζ = 0,0006
- Petersen's formula
for forcing function
2,0 - m = 13
0,0
1,65 1,75 1,85 1,95 2,05 2,15 2,25 2,35
excitation frequency (Hz)
7324
5388
15115
0,8 2663
15199
84% of pedestrians
acceleration (m/s²)
0,6
node 2663
node 15115
node 5388
0,4
node 7324
node 15199
- ζ = 0,0006
0,2
- Petersen's
formula for forcing
function
0,0
1,65 1,75 1,85 1,95 2,05 2,15 2,25 2,35
excitation frequency (Hz)
7 Verification criteria
Vibrations resulting from pedestrian excitation of a structure can lead to several forms
of distress:
- Intolerable vibration velocities and accelerations disturbing and discomforting
the users
- Overstressing of the structure
- Damage to non-structural elements
- Excessive noise (due to, for instance, reverberating equipment)
In most cases, only the first factor will have an influence on the design of the strcuture.
In order to verify that the structural response of the College Brug stands within the
acceptance limit of tolerance, it is important to express the relevant maximum values of
vibration amplitude and acceleration for the dynamic problem in order to establish a
comparison between these and the results of the study.
As it is shown in tables 10 and 11, two types of criteria may be adopted: one related to
human comfort and another related to lock-in effect.
Practice codes give more accurate data concerning this feature than the criteria found in
the literature references. For sake of clarity and simplicity, the British Standards seem to
be the most appropriate for the College Brug vertical vibration issue. According to the
BS 5400-2, the acceleration resulting from pedestrian excitation has to remain under a
certain threshold value given by the following formula:
a max,vertical ≤ 0.5 f 0
Lock-in effect
For this particular issue, Bachmann seems to be a reference. His studies demonstrate
that he deeply studied the problem of synchronisation between the pedestrians and the
vibrating structure. According to him, two criteria may be applied, one concerning
displacement and another concerning acceleration.
d max,vertical ≤ 10 mm
a max ,vertical ≤ 5 to 10%g ⇒ a max ,vertical ≤ 0.49 to 0.98m/s²
Concerning displacements, it is easy to note, looking at table 7 and figure 8 that the
amplitude never exceeds the threshold value given by Bachmann in order to avoid a
lock-in effect except for nodes 2663, 7324 and 15199 and for an excitation frequency
around 1,71Hz. Table 11 reports all values exceeding the threshold value of 10mm.
Table 11
Values of displacement exceeding the threshold value given by Bachmann (in mm)
fexc [Hz] node 2663 node 7324 node 15199
1.71 12.400 14.000 14.050
1.72 12.300 14.450 13.450
Concerning accelerations (figure 9), and looking at the criteria for human comfort given
in the British Standards BS 5400-2, results related to the same nodes 2663, 7324 and
15199 exceed the threshold value of 0.4458m/s²: These are reported in table 12.
Table 12
Values of acceleration exceeding the threshold value given in BS 5400-2 (in m/s²)
fexc [Hz] node 2663 node 7324 node 15199
1.68 / / 0.560
1.70 / / 0.575
1.71 0.780 0.850 0.870
1.72 0.760 0.890 0.930
1.75 / 0.540 0.470
Concerning the lock-in effect, Bachmann suggests for accelerations not to exceed a
threshold value lying between 0.49 to 0.98m/s². As it is possible to conclude from table
12, the values for acceleration concerning nodes 2663, 7324 and 15199 remain in the
range 0.56 to 0.93m/s² and never exceed the advised upper limit value of 0.98m/s².
However, as shown in table 11, displacements for frequencies 1.71 and 1.72Hz do
exceed the value of 10mm predicted by Bachmann as threshold level to avoid lock-in
effect. Indeed, the resulting values could generate a vibration that could bring the users
to get synchronised with the deck movements. Although concerned frequencies remain
quite low and are not susceptible to appear in a large number of cases, some solutions
could be brought in order to completely avoid a risk situation.
Several authors [1] [2] [3] [4] [5] propose some currently applicable remedial measures
in order to either avoid natural frequencies in the range 1.6 to 2.4Hz, increase damping
or limit vibration amplitudes by introducing external vibration absorbers.
8 References
[7] M. Willford, “Dynamic actions and reactions of pedestrians”, AFGC and OTUA
Footbridge Conference, Paris, 2002
[13] O.C. Zienkiewicz, R.L. Taylor, “The finite elements method, Volume 1: The
basis”, Butterworth-Heinemann, 2000
Table 11
Review of verification criteria appearing in codes of practice
Frequency
Name Application Criteria Comments
range
ISO 2631 Vertical and/or horizontal 1 to 80Hz Given in graphical form and expressed in relation to T = period of time (s) over which the acceleration
vibrations. Random, shock or effective acceleration: is measured
harmonic vibrations. T
1 2
a ( t ) dt=0,707 amax (m/s²)
T ∫0
aeff =
DIN 4150/2 Residential buildings. 1 to 80Hz Perception factor: d = displacement amplitude (mm)
f = basic vibration frequency (Hz)
0,8 f 2
KB=d (mm/s)
1+0,032 f 2 Maximum values of KB stand between 0,2 and 0,6
for continuous or repeated excitation.
KB is compared to reference values depending on vibration
frequency, duration of vibration, etc.
structures composed of 1, 2 or 3 Used to verify structure the following way : K = configuration factor
spans with constant cross-section. f0 > 5Hz Verification OK
B B
Ψ = dynamic response factor
f0 < 5Hz
B B amax,vertical ≤ 0,5 f0 at any part of
superstructure
Horizontal: amax,horizontal ≤ 1 to 2% g For Bachmann, acceptance criteria are related in most cases to physiological
effects on people representing serviceability problems, rather than safety problems
For avoiding lock-in effect: to the structure.
Vertical: d max,vertical ≤ 10mm For Bachmann, criteria are basically frequency-dependent
Michael Wilford For pedestrian comfort: a = acceleration resulting from the structure’s vibration
(reference [7]) Vertical: amax,vertical ≤ 7% g
Wilford bases his theory on measurements that took place on the London
Horizontal: amax,horizontal ≤ 0, 2% g Millennium Bridge.
For avoiding lock-in effect:
Vertical: amax,vertical ≤ 4% g
Horizontal: amax,horizontal ≤ 0, 25% g
Stoyan Stoyannoff For human comfort: a = acceleration resulting from the structure’s vibration
(reference [11]) Vertical: amax,vertical ≤ 0,07 g
According to Stoyannoff, people can tolerate different levels of vibration
Horizontal: amax,horizontal ≤ 0,02 g depending on activity. Acceptable levels for people working in offices will differ
significantly from values for people participating in rhythmic activities.