Sei sulla pagina 1di 16

Vibration serviceability of footbridges:

a comparative study of 3 design methodologies


K. Van Nimmen1, P. Van den Broeck2, B. Gezels2 & G. De Roeck1
1
K.U.Leuven, Departement Burgerlijke Bouwkunde
Kasteelpark Arenberg 40, B-3001, Heverlee, Belgium
2

KaHo St.-Lieven, Departement Industrieel Ingenieur, afdeling Bouwkunde-Landmeten,


Gebroeders Desmetstraat 1, B-9000, Gent, Belgium

Abstract
Vibration serviceability has become an important issue in the design of modern slender footbridges with
large spans. The natural frequencies of these bridges are often in the range of the loading frequencies of
the dynamic walking excitation of (groups of) pedestrians. This causes vibrations which may reduce the
comfort of the users of the footbridge. This paper presents the evaluation of the vibration serviceability of
a footbridge in Wetteren according to 3 recent design methodologies: Willford M.R. & Young P. (2006)
[1], the French guideline of Charles P. & Hoorpah W. (2006) [2] and the European guideline Hivoss
(2008) [3]. These design procedures are compared regarding the modeling and the assumptions made on
the following issues: the modeling of the human-induced loads, the characterization of the dynamic
behavior of the footbridge, the calculation of the vibration levels and the evaluation of these vibration
levels. The results illustrate the different approaches used in the design procedures leading to a different
evaluation of the vibration serviceability of the footbridge.

Introduction

The trend towards lightweight and slender structures increases the susceptibility to dynamic loads. For
footbridges this may lead to unacceptable vibrations due to the walking excitation of (groups of)
pedestrians. Famous examples of these vibration serviceability problems are the Millennium bridge in
London and the Solferino footbridge in Paris with unexpected lateral vibrations due to a lock-in
phenomenon. These events have initiated research and the publication of design guidelines to check the
dynamic behavior of footbridges. Nevertheless more research is still needed for the modeling of humaninduced vibrations due to complexity of the stochastic pedestrian load.
In this paper, the vibration serviceability of a footbridge in Wetteren is evaluated according to 3 recent
design methodologies: Willford M.R. & Young P. (2006) [1], A Design Guide for Footfall Induced
Vibration of Structures, The Concrete Center, the French guideline of Charles P. & Hoorpah W. (2006)
[2], valuation du comportement vibratoire des passerelles pitonnes sous laction des pitons,
Association Franaise de Gnie Civil, Stra/AFGC, France and the European guideline Hivoss (2008) [3],
Design of footbridges HIVOSS (Human Induced Vibrations of Steel Structures ) which is based on the
results of a European research project [4]. These design procedures are compared regarding the modeling
and the assumptions made on the following issues: the modeling of the human-induced loads (single
pedestrian versus groups of pedestrians with varying pedestrian densities and synchronization effects,
deterministic versus stochastic), the characterization of the dynamic behavior of the footbridge (estimation
of the modal parameters: natural frequencies, mode shapes and damping factors), the calculation of the
vibration levels (single mode versus modal superposition, directions: vertical, lateral and longitudinal) and
the evaluation of the vibration levels (weighing factors, vibration criteria). After the discussion of the
methodologies, the results of the vibration serviceability check of the footbridge in Wetteren will be
compared.

1491

1492

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

For this footbridge, extensive experimental and numerical results are available: a series of vibration
measurements with different-sized groups of pedestrians crossing the bridge for the case of free and
synchronized walking and the data of an operational modal analysis used for the updating of a finite
element model of the footbridge. The availability of measured damping ratios and an updated finite
element model provides the possibility to perform a comparative study with the proposed damping ratios
and the initial finite element model respectively, to study the influence on the evaluation of the vibration
serviceability of the footbridge according to the investigated design methodologies.

Wetteren footbridge

The footbridge, over the E40 highway (Brussels-Ghent, Belgium), provides a connection for cyclists and
pedestrians between Wetteren and Zottegem (Figure 1). The steel bridge has two spans: a large span of
75.23m, which is designed as a bowstring bridge with
outward inclined bows (13.78 and a maximum height
of 12.82m with respect to the bridge deck), and a short
span of 30.33m. The two main beams along the bridge
length, at each side of the bridge deck, have a distance
of 3.55m between the axes. The bridge has been subject
of previous research. An extensive measurement
campaign was carried out to derive the modal
parameters which were then used to develop an updated
finite element model as described by Degrauwe et al.
Figure 1: Wetteren footbridge over the E40 [5]. Two types of updating variables were considered in
this analysis: boundary conditions and cable stiffnesses.
Highway (Belgium)
At the midspan support, the bows and support legs are
welded to a stiff connector that is bolted to a concrete foundation. This support is modeled with two
rotational springs, one in the bridge axis direction, the other in the lateral direction (assuming that both
strings have the same stiffness, Km ). The other bow ends and the bridgedeck support are implemented
with neoprene supports, modeled with three springs (one in each direction: K n,x , K n,y , K n,z ) at each
support. The respective springs stiffnesses and the reduction factor for the cable stiffnesses , obtained
from the updating procedure, are presented in Table 1. Table 2 presents the derived modal parameters as
well as the calculated natural frequencies of interest, in the range up to 5Hz (susceptible to human-induced
vibrations), and the corresponding measured damping ratios.
Another research included the measurement of the vertical and the lateral accelerations at seven locations
along the bridge length (Figure 2) with different-sized groups of pedestrians crossing the bridge, varying
in number from 10 up to 50 participants, for the case of free walking and synchronized as described by
Van den Broeck et al. [6]. The maximum RMS-values of the vertical acceleration (mean value for the
passages) are shown in Figure 3 as a function of the position x along the bridge length for the free and the
synchronized walking experiments. Also shown in Figure 3 are the ambient acceleration levels at the
measurement locations (characterized by the maximum RMS-value of the vertical acceleration) mainly
due to the wind and the passing traffic on the highway.

Km [x109 Nm/rad]
K n,x [x109 N/m]
K n ,y [x109 N/m]
K n,z [x109 N/m]
[-]

Initial
model

Updated
model
2.196

1.043

0.064

27.139

1.0

0.961

Table 1: Results of the updating procedure

Figure 2: Measurement locations along the bridge length

DYNAMICS OF BRIDGES

No.

1493

Mode shape

Measurements

Finite element model


Initial

Updated

fmeas , j [Hz]

meas, j [%]

ffem, j [Hz]

ffem, j [Hz]

0.71

2.12

0.72

0.70

1.67

0.21

1.71

1.65

2.19

0.55

2.36

2.25

3.74

0.76

3.77

3.74

3.93

3.88

4.00

3.90

4.19

4.09

4.44

0.56

4.74

4.41

Table 2: Natural frequencies fmeas , j and damping ratios meas, j derived from measurements, natural frequencies ffem, j
calculated via the initial and updated finite element model

(a)

(b)

Figure 3: Maximum RMS-value of the vertical acceleration at the measurement locations during the passage of 10 (G10 - ), 20
(G20 - *), 30 (G30 - ), 40 (G40 - +) and 50 (G50 - ) students in the case of (a) free walking and (b) synchronized walking. The
ambient vibration levels, derived from a measurement period of 13 minutes, are also indicated (Amb - ).

1494

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

Methodology of Willford & Young, Setra and Hivoss design guide

This section gives a summary of the design procedures described by Willford & Young [1], the French
Setra guideline [2] and the European Hivoss guideline [3]. Due to their big similarity, Setra and Hivoss
are presented in a joint summary (per subsection).

3.1

Modeling the human-induced loads

3.1.1

Setra and Hivoss

Both guides specify a classification for the expected pedestrian traffic. The Setra guideline defines 4
classes: Class IV - low usage footbridge, Class III - normal usage footbridge, Class II - urban footbridge,
linking populated zones and Class I - urban footbridge, high concentration of pedestrian traffic or with
crowds frequently crossing the structure. The Hivoss guideline specifies 5 traffic classes varying from
TC1 - very weak, TC2 - weak, TC3 - dense, TC4 - very dense to TC5 - exceptional dense traffic. The
corresponding pedestrian densities d [persons/m2] can be found at the top of Table 8. The selected
pedestrian density results into a random stream of N pedestrians.
The key element in both Setra and Hivoss is to develop a practical design procedure, transforming the
complex stochastic load due to a stream of N random pedestrians, into an equivalent deterministic
uniformly distributed harmonic load (simulating an equivalent number of Neq perfectly synchronized
pedestrians) which generates the same acceleration as the 95% percentile-value of the peak accelerations
due to the random stream. The amplitude qeq,e of the equivalent load, in direction e (vertical, lateral or
longitudinal) is equal to:
qeq,e

Neq
S

ehG eh (f j )

[N/m]

(1)

where subscript j denotes the mode shape under consideration, S is the bridge deck surface area,
N eq 10.8 j N
N eq 1.85

(j

damping ratio, section 3.2.1) for pedestrian densities

d 1 pers/m and

N for d 1 pers/m , eh is the dynamic load factor of the h-th harmonic of the load in

direction e generated by a single pedestrian with body weight G (Table 3) and eh (f j ) is the reduction
coefficient taking into account the probability of footbridge resonance occurring depending on the value of
the natural frequency f j with respect to the range of possible step frequencies. As can be observed from
Table 4, slightly different coefficients are used by both design guides for the vertical and longitudinal
direction. Also for the lateral direction, differences are present and no second harmonic of the pedestrian
load in the lateral direction is considered by the Hivoss guide. The sign of the equivalent load at position p
on the bridge deck, follows the sign of the modal deformations

Direction e
Vertical
Lateral
Longitudinal

j, ep

for each investigated mode shape .


j

h=1
[-]
0.4
0.05
0.2

h=2
G [N]
280
35
140

[-]
0.1
0.01
0.05

G [N]
70
7
35

Table 3: Dynamic load factor eh of the h-th harmonic of the load generated by a single pedestrian with body weight
G = 700 N and corresponding force amplitude eh G .

DYNAMICS OF BRIDGES

1495
vertical ,longitudinal

<1
Setra
Range

1.7
2

2.1
1

2.6
2

1.25

4.2

2.3 2.5

Hivoss
Range

3.4

3.4

>5
4

4.2

4.6

lateral

< 0.3
Setra
Range

0.3

0.5

1.1

1
0.5

Hivoss
Range

1.3
2

0.7

1.7

2.1
3

2.5

>2.5
4

1.2

Table 4: Reduction coefficient

eh

( f ) and frequency ranges for the vertical, longitudinal and lateral vibrations,
j

according to the Setra() and the Hivoss (--) design guide.

3.1.2

Willford & Young

The method presented in this guideline is based on the loading resulting from one single pedestrian,
leaving out of consideration the complexity of groups of pedestrians. The application area of this guideline
is spread over floors and footbridges, making a distinction between high and low frequency structures,
resulting in different loading models (impulsive and harmonic respectively). Since footbridges can be
considered to be low frequency structures, only the corresponding harmonic loading model is considered
in this paper.

1496

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

In this guideline, only the vertical component of the pedestrian load that is taken into account, lateral and
longitudinal vibrations are left out of consideration.
As a first parameter characterizing the pedestrian load, this guideline proposes maximum footfall rates for
different environments. Concerning footbridges, this maximum rate is set to 2.5 footfalls per second. This
leads towards the following interval of step frequencies that needs to be considered: 1- 2.5 Hz.
The pedestrian load considers 4 harmonics, each one accompanied by a dynamic load factor (dynamic
force expressed as a fraction of a persons static weight) (Table 5). The stochastic character of the load is
translated into mean and design values, each corresponding to a likelihood of exceedence, design
values having a 25% chance of exceedence. The amplitude of the harmonic force can then be found as
Fh h .G , with G the weight of the pedestrian.
Harmonic
number h
1
2
3
4
h>4

Harmonic forcing
frequency fh [Hz]
1 2.8
2 5.6
3 8.4
4 11.2
> 11.2

Coefficient of
variation Cv
0.17
0.40
0.40
0.40

Dynamic load factor h


Mean value
Design value
0.37 ( fh 0.95) 0.5
0.054 + 0.0044 fh
0.026 + 0.0050 fh
0.010 + 0.0051 fh
0

0.41 ( fh 0.95 ) 0.56


0.069 + 0.0056 fh
0.033 + 0.0064 fh
0.013 + 0.0065 fh
0

Table 5: Average and design values and coefficients of variation of the footfall harmonics (according to [1])

3.2

Characterization of the dynamic behavior of the footbridge

3.2.1

Setra and Hivoss

Damping ratios, necessary for the calculation, are suggested for different construction materials in the two
guides. In case of a steel footbridge, both suggest an average damping ratio of 0.4%.
The procedures require all modes with frequencies up to 5 Hz. The risk of footbridge resonance is
determined by an evaluation of the natural frequencies in relation to the possible loading frequencies
present in the pedestrian load. The Setra guideline uses 4 ranges to characterize this risk: Range 4 negligible risk (no calculation required), Range 3 - low risk of resonance with the 2nd harmonic, Range 2 medium risk and Range 1 - maximum risk of resonance with the 1st harmonic. The Hivoss guideline
specifies 2 critical ranges for which a calculation is required: Range 1 - risk of resonance with the 1st
harmonic and Range 2 - risk of resonance with the 2nd harmonic. The corresponding values of the natural
frequencies can be found in Table 4. When the considered frequency is classified in range 2 or 3 in
combination with a bridge class III, or overall classified in range 4, the demand of a dynamic calculation is
stated to be unnecessary according to Setra.
The variation of the natural frequency and corresponding mode shapes due to the mass of the pedestrians
on the bridge deck is taken into account depending on the pedestrian density. The Hivoss guideline
however states that this influence is only to be taken into account when the mass of the pedestrians
exceeds 5% of the modal mass of the bridge deck (of the considered mode).
3.2.2

Willford & Young

The guideline suggests damping ratios for different construction types, in case of steel bridges, this ratio is
0.3-0.5%.
The procedure presented to calculate the resonant response, requires all vertical modes with frequencies
up to 15 Hz. These modes potentially have a significant contribution to the response and should be
included in the calculations by means of modal superposition.

DYNAMICS OF BRIDGES

1497

3.3

Calculation of the vibration levels

3.3.1

Setra and Hivoss

Since for the calculation of the maximum acceleration u j ,e max of the bridge deck resonance conditions are
considered and based on a discretisation Si of the bridge deck surface area ( Si Si ) with the
i

corresponding positions p(i ) , the maximum acceleration can be calculated as (with mass normalized mode
shapes):
u j ,e

max

Si j , pe

eq ,e

max j , p( i )e
p( i )

(2)

with: u j ,e max maximum acceleration in direction e [m/s], qeq,e amplitude of the equivalent load, in
direction e [N/m], Si discretisation of the bridge deck ( S Si ) [m], p(i ) position on the bridge deck
i

(according to the discretisation) [-], j ,pe modal displacements of mode j, in position p(i ) and direction e
1 / kg and j damping value for the considered mode j [-].

3.3.2

Willford & Young

The response is dominated by resonance response, resulting from the contributions of the four harmonics.
To limit the number of calculations, the critical step frequencies ( fc1 fcn ) are calculated as fractions of the
natural frequencies (1, 1/2, 1/3, 1/4). Each fraction that lies within the investigated interval of step
frequencies (3.1.2), as well as the limits of the interval, are considered to be critical step frequencies for
which the total response will be calculated. The total response can be found performing modal
superposition for all modes up to 15 Hz.
The maximum acceleration, per critical step frequency, results from the most critical combination of input
(position of the walking force) and output (position response) point. The total response (for all modes) per
harmonic equals the product of the transfer function ( Hhu, input output ) with the amplitude of the
corresponding harmonic force ( Fh ):
uh Fh Hhu, input output

with h 1, , 4

(3)

For each critical step frequency and its harmonics, these acceleration levels are calculated for each
combination of input and output point, and will be assessed in the next phase.

3.4

Evaluation of the vibration levels

3.4.1

Setra and Hivoss

The assessment of the comfort level, according to the Setra and the Hivoss guideline is based on Table 6
for the vertical and the horizontal (lateral and longitudinal) vibrations. The comfort level is determined for
each mode separately, indicating critical modes.
Lateral acceleration levels of 0.1 to 0.15m/s2 can trigger the lock-in phenomenon. Below this threshold,
pedestrian behavior can be considered to be random. As soon as the threshold is passed, the rate of
synchronization rises significantly as well as the acceleration level, where it becomes uncomfortable.

1498

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010


Acceleration [m/s]

Vertical
Setra
Hivoss

<0.5

0.5

max

Horizontal
Setra
Hivoss

max
max

2.5

mean

0.15
0.3
mean
mean

min

0.8

>2.5
unacc

> 0.8
unacc
unacc

min
min

0.1

Table 6: Comfort levels and corresponding acceleration ranges for vertical and horizontal vibrations

3.4.2

Willford & Young

This guideline expresses the vibration performance, in terms of RMS acceleration and response factors.
For the head-to-toe axis (z), the acceleration level corresponding to the average threshold of perception
uR 1 , is plotted in Figure 4. The presented curve is determined by the following equations (accelerations
in m/s):
fh 4 Hz

uR 1,h

0.0141
fh

4 Hz fh 8 Hz

uR 1,h 0.0071

fh 8 Hz

uR 1,h 2.82 fh x 104

The response factor R is simply a multiplier on the level of vibration at the average threshold. For each
computed acceleration, the corresponding R-factor is calculated:
Rh

uh
uR 1,h

The resulting response factor, for the considered critical step


frequency: Rf Rci2 1 Rci2 2 Rci2 3 Rci2 4
ci

The final response factor is defined as the overall maximum for


all combinations of input and output positions and over all
considered critical step frequencies: R max Rf , Rf ... Rf
c1

c2

cn

with fc1 fcn all critical step frequencies.


The guide refers to (BS 5400, 1978) [9] providing guidance on
the acceptable levels of vertical vibration for footbridges. They
present a refinement of these values, for single person excitation
at the most critical footfall rate (Table 7). These limits are
however found unnecessarily severe ([7], [8]) in case of floors.

External bridges
Indoor bridges
Indoor bridges that are not particularly lightweight, or are
exposed, high in an atrium or heavily trafficked

Table 7: R-values - bridges

Figure 4: Threshold of perception

R < 64
R < 32
R < 24

DYNAMICS OF BRIDGES

1499

For each critical step frequency (and each combination of input-output), the RMS acceleration is also
calculated (with h the number of the harmonic):
4

uRMS

3.5

u
h 1

2
h

Similarities and differences

Setra and Hivoss translate the loading of a group of pedestrians to an equivalent load of Neq pedestrians,
synchronized with the considered natural frequency of the footbridge, uniformly spread across the bridge
deck, generating the same maximum acceleration as the random group of N pedestrians. The pedestrian
loading presented consists of two harmonics. The loading proposed by Setra and Hivoss is identical, apart
from a small difference in reduction factors (Table 4). Regarding the lateral vibrations, the Hivoss
guideline does not examine the response due to the second harmonic. As a result of the resonance response
method, each vibration mode of the footbridge is separately assessed (single mode assessment).
Setra and Hivoss enable the evaluation of the footbridge for different design situations, resulting into
different pedestrian densities and the required comfort level.
Willford & Young do not consider the loading of groups of pedestrians, this guideline presents the loading
of one single pedestrian making it less suitable for the design of footbridges. An interval of step
frequencies is considered using four harmonics. Critical conditions (step frequencies) are identified at
which resonance could occur, leading to high vibration levels. In this guide, the vertical component is the
only component of the pedestrian load that is taken into account, lateral and longitudinal vibrations are left
out of consideration.
Willford & Young apply the concept of modal superposition and the assessment results into one single
response factor which has to be evaluated. No different design situations are considered.
The variation of the natural frequency and corresponding mode shapes due to the mass of the pedestrians
on the bridge deck is taken into account by Setra and Hivoss. This influence is not considered in the
guideline of Willford & Young, notwithstanding its significant impact in case of high pedestrian densities.

Results

This section presents a study of the maximum acceleration and the corresponding evaluation as predicted
by the Willford & Young and the Setra and the Hivoss design guide for different values of the pedestrian
density.

4.1

Vertical accelerations

4.1.1

Setra and Hivoss

For each mode shape of the updated finite element model, with a natural frequency < 5Hz, the maximum
vertical acceleration and corresponding comfort level, according to the Setra and the Hivoss design guides,
are summarized in Table 8 for a range of pedestrian densities varying from 15 persons/deck to 1.5
persons/m2. Only the results are given for which a calculation is required according to the design
procedures of both guides. All mode shapes and frequencies are calculated taking into account the mass of
the pedestrians present on the bridge deck.
A number of interesting observations can be made. The variation of the natural frequency with increasing
pedestrian density, due to the extra mass on the bridge deck (estimated as 70 kg/person), is quite different

1500

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

for each mode shape. A substantial variation, as can be observed for mode shapes 3, 4 and 5, indicates a
sensitivity of the mode shape for the pedestrian load due to important modal deformations of the bridge
deck (Table 2). The investigation of the comfort levels enables to identify the critical vibration modes for
which the comfort may decrease to a mean, minimum or even unacceptable level with increasing
pedestrian density. For the case of the Wetteren footbridge, the fourth mode could be critical, due to two
reasons: a natural frequency in the range of the 2nd harmonic loading frequency of the pedestrian load and
an important deformation of the short span as can be observed in Table 2. It should be noted that the
predicted maximum acceleration levels are very sensitive (inversely proportional) to the value of the
damping ratio used in the calculations. In Table 8, the assumed damping ratio is = 0.4 %, as suggested in
both guides for the case of steel bridges. The comparison with the measured values of the Wetteren
footbridge (Table 2) indicates that this is a conservative value for most mode shapes, with an exception for
mode shape 2. The use of the measured damping ratio can have a significant influence on the predicted
acceleration level and the corresponding comfort level as can be illustrated for the fourth mode shape: for
a pedestrian density equal to 1 p/m2, the evaluation of umax =3.2m/s2 and unacceptable comfort level, based
on = 0.4%, changes to umax =1.69m/s2 and minimum comfort level, based on the measured value =
0.76%.

Comparing the results of the Setra and the Hivoss design guide, it can be observed that the predicted
acceleration levels are quasi identical where the small differences can be explained by the difference of the
used eh -functions in the design guides. An important difference however, can be observed for the fourth
mode shape and a pedestrian density equal to 0.5 p/m2. According to the Setra design guide this
corresponds to Class III and a low risk of resonance (Range 3) with the second harmonic of the pedestrian
load, for which no calculation is required. The results of the Hivoss guide however indicate that this
situation may lead to a mean comfort level and can be important for the check of the vibration
serviceability.

(p=person)

15
[p/deck]

0.2
[p/m2]

15
2.6

68
5.6

TC 1

TC 2

0.7

0.7

N
Neq
Setra
Hivoss
f1 [Hz]

Pedestrian density d
0.5
0.8
[p/m2]
[p/m2]
170
9
Class III
TC3

271
11
Class II

0.7

Setra

1
[p/m2]

1.5
[p/m2]

339
34
Class I
TC 4

509
42

0.7
-

TC 5

0.7
-

0.7
-

max [m/s2] - CL
Hivoss

f2 [Hz]

1.65

1.65

1.64

1.63

Setra

0.04

1.62
max

1.6

0.12

max

0.11

max

max [m/s2] - CL
Hivoss

0.01

f3 [Hz]

max

0.02

2.24

max

0.03

2.21

max
2.17

Setra

2.14

0.05

max

0.09

0.14

2.11
max

max
2.06

0.27

max

0.26

max

max [m/s2] - CL
Hivoss

0.01

f4 [Hz]

max

0.02

3.74

max
3.67

3.55

Setra

3.44

0.95

mean

1.13

0.29

3.38
min

max
3.22

3.2

unacc

3.2

unacc

max [m/s2] - CL
Hivoss

0.31

f5 [Hz]

max

0.64

3.87

mean
3.84

3.79

Setra

3.74

0.03

max

0.04

2.97

3.71
max

unacc
3.63

0.11

max

0.11

max

max [m/s2] - CL
Hivoss

0.01

f6 [Hz]

max

0.02

3.90

max
3.90

3.89

Setra

3.88

0.05

max

0.16

0.13

3.87
max

max
3.80

0.83

mean

0.83

mean

max [m/s2] - CL
Hivoss

0.005

max

0.02

max

1.58

min

DYNAMICS OF BRIDGES

f7 [Hz]

1501

4.08

4.05

4.00

Setra

3.96
-

0.33

3.94
max

0.64

3.92
mean

max [m/s2] - CL
Hivoss

0.11

f8 [Hz]

max

0.23

4.40

max

0.34

4.39

max

0.64

4.36

Setra

4.33

0.01

max

0.02

mean

0.19

4.32
max

max
4.28

0.05

max

0.04

max

max [m/s2] - CL
Hivoss

0.003

max

0.01

max

0.06

max

Table 8: Maximum vertical acceleration and corresponding comfort level CL, calculated for the natural frequencies f 1 - f8 < 5Hz
of the updated finite element model, as a function of the pedestrian density d, according to the Setra and the Hivoss design guide,
based on the suggested damping ratio for steel bridges = 0.4 %. N (Neq) corresponds to the number of (equivalent) pedestrians.
Empty cells indicate d is not considered by the guide and - indicates that no calculation is required according to the guide.

4.1.2

Willford & Young

The maximum RMS acceleration levels and response factors for the updated and the initial finite element
model are summarized in Table 9. The maximum response factor in each output point on the bridge deck
(for the updated model), is presented in Figure 5. The overall maximum response factor (107) is the global
response factor that needs to be evaluated. This response factor is the result of the critical step frequency
of 1.87 Hz (= f4 /2) with the pedestrian positioned in the center of the short span. In other words, this
maximum is mainly the result of resonance of the second harmonic with the fourth natural frequency. The
corresponding mode shape, characterized by bending deformation in the short span, can clearly be
identified in Figure 5. This R-value
exceeds the limit of 64 posed by
Willford & Young. According to this
guideline, the evaluation of the dynamic
behavior of the concerned footbridge is
negative. However, as mentioned before,
these limits are found unnecessarily
severe ([7], [8]).
To have an indication of the comfort
level corresponding to the RMS
acceleration as calculated according to
Willford & Young, a corresponding
approximation
of
the
maximum
acceleration is calculated ( umax , Table
9). This maximum acceleration can then
be evaluated using the criteria of Setra
and Hivoss. According to Setra and
Hivoss, the footbridge meets the criteria
of mean comfort (Table 9).

Response factor R [-]

Figure 5: R-factor for different output points along the


bridge length design; updated model

uRMS ,max [m/s]

Equivalent assessment according


to Setra and Hivoss (design)

Mean

Design

Mean

Design

umax [m/s]

Comfort level

Updated model

83

106

0.43

0.54

0.76

mean

Initial model

84

107

0.43

0.54

0.76

mean

Table 9: Maximum RMS acceleration ( uRMS ,max ) and maximum response factor according to Willford & Young, mean and design
loading, for the updated and initial finite element model

1502

4.2

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

Lateral accelerations

For each mode shape of the updated finite element model, with a natural frequency < 2.5Hz, the maximum
lateral acceleration and corresponding comfort level, according to the Setra and the Hivoss design guides,
are summarized in Table 10 for a range of pedestrian densities varying from 15 persons/deck to 1.5
persons/m2. All mode shapes and frequencies are calculated taking account of the pedestrian density
present on the bridge deck. The effect of lock-in has to be checked for mode shapes with lateral modal
deformations. No evaluation for lateral accelerations is considered by Willford & Young.
An important difference between the Setra and the Hivoss design guide, is that no effect of the second
harmonic of the pedestrian load is to be considered according to the Hivoss design guide while this effect
is only disregarded by the Setra design guide for a pedestrian density of Class III. This difference is
extremely important for the evaluation of the lateral acceleration levels: while the behavior is satisfactory
according to the Hivoss design guide, the Setra design guide indicates a risk of lock-in for the third mode
shape due to the effect of second harmonic of the pedestrian load.

(p=person)
Setra
Hivoss
N
Neq

15
[p/deck]

0.2
[p/m2]

TC 1
15
2.6

TC 2
68
5.6

0.7

0.7

f1 [Hz]
Setra
max [m/s2] - CL
Hivoss

0.003

max

0.006

1.65

max

0.01

1.65

2.24

0.01

1.5
[p/m2]

0.7

0.7

0.04

0.02

max

2.14
-

0.05

max

0.05

1.62

TC 5
509
42

max

0.04
1.63

2.17

max

max
1.64

2.21

1
[p/m2]
Class I
TC 4
339
34

0.7
max

f3 [Hz]
Setra
max [m/s2] - CL
Hivoss

0.7
0.01

f2 [Hz]
Setra
max [m/s2] - CL
Hivoss

Pedestrian density d
0.5
0.8
[p/m2]
[p/m2]
Class III
Class II
TC3
170
271
9
11

max

0.06

max

2.11
0.17 lock-in
-

max
1.60

2.06

Table 10: Maximum lateral acceleration and corresponding comfort level CL, calculated for the natural frequencies f 1 - f3 < 2.5Hz
of the updated finite element model, as a function of the pedestrian density d, according to the Setra and the Hivoss design guide,
based on the suggested damping ratio for steel bridges = 0.4 %. N (Neq) corresponds to the number of (equivalent) pedestrians.
Empty cells indicate d is not considered by the guide and - indicates that no calculation is required according to the guide.

4.3

Comparison of the results

Figure 6 and Figure 7 present the comparison of the results for the vertical vibrations between Setra and
Hivoss, respectively for the 4th and the 7th mode for the different pedestrian densities, as well as the
comparison to the global maximum acceleration level calculated according to Willford & Young (Table 9)
generated by one single pedestrian. Comparing the results between Setra-Hivoss, and Willford & Young,
is not easy because of the inherent differences in methodology.

DYNAMICS OF BRIDGES

1503

In addition to the results from the


different
guidelines,
the
comparison is made between the
results from the initial and the
updated finite element model on
one hand,
and between the
damping ratios proposed by the
guides and measured damping
ratios (for the updated model) on
the other hand.

Figure 6: Vertical acceleration level according to Young (---- max); Vertical


acceleration for different pedestrian densities for mode 4 according to Setra and
Hivoss: Setra ( updated model, initial model), Hivoss ( updated model, +
initial model), updated model and measured damping ratio.

A
number
of
interesting
observations can be made.
Comparing the results from the
different
guidelines,
a
considerable agreement is found
between Setra and Hivoss, as
previously explained in (4.1.1).
The restriction of Willford &
Young making no distinction for
different bridge classes and
pedestrian densities, is also
illustrated by these figures.
Table 9 illustrates that the
difference between the results of
the updated and
the
initial
finite element model according
to Willford & Young is
negligible. This is due to the fact
that the difference in frequency
between the models for the
dominant mode (f4) is small.
Considering the results of the 4th
mode (Figure 6), only little
differences arise between the
results for the updated and the
initial model. On the other hand
for the 7th mode (Figure 7), the
difference in model results in a
different comfort class for high

Figure 7: Vertical acceleration level according to Young (---- max); Vertical


acceleration for different pedestrian densities for mode 4 according to Setra and
Hivoss: Setra ( updated model, initial model), Hivoss ( updated model, +
initial model)

pedestrian densities (>1p/m).


This
should
indicate
the
importance of a reliable (well
tuned) finite element model. This
will always require the estimation
of a number of uncertain
parameters as for example the
support conditions of the bridge.
In this case, the differences are
rather small, however, one must
be conscious of the uncertainties
and take them into account when
assessing the bridges behavior.

1504

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

Since the 7th mode shape could not be paired to a measured one, the influence of damping ratio is only
made for the 4th mode shape (Figure 6). It is clear that the damping ratio has an important influence on the
predictions, in the present case, it results in different comfort classes. However, this factor can only be
estimated in the design phase.
A comparison of the results for lateral and longitudinal vibrations, is not possible because Willford &
Young does not imply calculation of vibration levels in these directions. This is considered a drawback of
this method towards modern footbridges.
In Figure 6, a shaded area can be noticed between 0 0.2 p/m with a maximum height of about 0.8 m/s,
corresponding to the magnitude of the acceleration level measured at the short span (Figure 3). This
pedestrian density more or less corresponds to the population on the bridge deck during the measurement
campaigns [6]. This illustrates that the predicted vibration levels of the guides are in the same order of
magnitude of the measured acceleration levels.
Setra and Hivoss apply single mode assessment resulting into a detailed overview of the importance of
each mode. This enables the designer to provide the construction of possible specific adjustments (for
instance, installation of TMDs). When applying Willford & Young, the same critical vertical mode shape
is clearly identified and the method modal superposition ensures in this case the contribution of different
modes to the global result.

Conclusions

The presented results demonstrate that the proposed methods enable the designer to check the vibration
serviceability of the footbridge based on a prediction of the maximum acceleration and the corresponding
comfort evaluation. Both Setra and Hivoss guidelines anticipate on different pedestrian densities. It is the
decision of the bridge owner to evaluate whether the different design situations (combinations of
pedestrian density and required comfort level) correspond to acceptable vibration levels. Currently, no
codes exist where similar requirements are implemented.
For the case of the Wetteren footbridge, the assessment is negative according to Willford & Young. The
daily conditions of use of this footbridge can be characterized as normal, with sometimes crossing of large
groups (due to a school situated nearby). This pedestrian density can be classified as Class III according to
the Setra design guide and as TC 3 according to the Hivoss design guide. The calculated results indicate
that the maximum comfort level may be expected in these conditions.
Although the proposed design procedures in the guides provide very useful results for design purposes, it
is important to keep in mind that the methods are based on the key assumption that the effect of the
complex stochastic dynamic load due to a stream of pedestrians may be estimated by the effect of a
deterministic load of an equivalent number of stationary pedestrians. The method proposed by Willford &
Young is based on the loading model of one single pedestrian, moreover, only considering its vertical
component, making it less suitable for the design of (urban) footbridges.
All three methods rely on the availability of accurate modal parameters of the bridge. The damping ratio
has an important influence on the predictions but can only be estimated in the design phase. Based on a
finite element model the natural frequencies and mode shapes can be calculated but this requires the
estimation of a number of uncertain parameters as for example the support conditions of the bridge.
As a more general conclusion, in order to obtain a reliable design methodology more fundamental research
is required for the modeling of the human-induced forces due to groups of pedestrians, including both
synchronization effects (between the pedestrians themselves and between the pedestrians and the
footbridge) as the effect of the pedestrians on the modal parameters of the footbridge.

DYNAMICS OF BRIDGES

1505

Acknowledgements
These results were partly obtained within the framework of a current research project, TRICON [10],
financed by the Flemish government (IWT, agency for Innovation by Science and Technology).

References
[1] Willford, M. R., & Young, P. (2006). A Design Guide for Footfall Induced Vibration of Structures.
Camberley, Surrey: The Concrete Center.
[2] Charles P. & Hoorpah W., valuation du comportement vibratoire des passerelles pitonnes sous
laction des pitons, Association Franaise de Gnie Civil, Stra/AFGC, France, 2006.
[3] Design of footbridges HIVOSS (Human Induced Vibrations of Steel Structures), 2008.
[4] Butz C. et al., Advanced load models for synchronous pedestrian excitation and optimised design
guidelines for steel footbridges (SYNPEX), final report, Research Fund for Coal and Steel, 2007.
[5] Degrauwe D., Reynders E., De Roeck G. and Van den Broeck P., Operational modal analysis and
updating of a footbridge, In Proceedings of the Footbridge 2008 Conference, Porto, Portugal, July
2008.
[6] Van den Broeck P., De Roeck G., Reynders E. and Degrauwe D., Experimental and numerical
analysis of the pedestrian-induced vibrations of a footbridge, In Proceedings of the Footbridge 2008
Conference, Porto, Portugal, July 2008.
[7] European Commission - Technical Steel Research. (2006). Generalisation of criteria for floor
vibrations for industrial, office, residential and public building and gymnastic halls. RFCS Report
EUR 21972 EN.
[8] Floor vibrations HIVOSS (Human Induced Vibrations of Steel Structures), 2008.
[9] BS 5400, 7. (1978). Steel, concrete and composite bridges. Specification for materials and
workmanship, concrete, reinforcement and prestressing tendons.
[10] TRICON, Van den Broeck P., De Roeck G., IWT project, Prognose en controle van mensgenduceerde trillingen in bouwkundige constructies (Prediction and control of human-induced
vibrations of civil engineering structures).

1506

P ROCEEDINGS OF ISMA2010 INCLUDING USD2010

Potrebbero piacerti anche