You are on page 1of 10

Experimental Vibration Analysis for Civil Engineering Structures

189

Experimental validation and calibration of pedestrian loading


models for footbridges
F. Ricciardelli
University of Reggio Calabria, Reggio Calabria, Italy

C. Briatico
University of Florence, Florence, Italy

E.T. Ingolfsson & C.T. Georgakis


Danish Technical University, Lyngby, Denmark

ABSTRACT: Different patterns of pedestrian loading of footbridges exist, whose occurrence


depends on a number of parameters, such as the bridge span, frequency, damping and mass, and
the pedestrian density and activity. In this paper analytical models for the transient action of one
walker and for the stationary action of an incoherent crowd of walkers will be presented, whose
parameters have to be calibrated based on experimental tests. Results will be summarised of
tests aimed at a statistical characterisation of the walking parameters and of the walking forces.
In addition, results of tests performed on a model footbridge will be shown, aimed at setting the
limits of applicability of the proposed models.
1 INTRODUCTION
Footbridge response to pedestrian loading is known to be strongly related to the possibility of
occurrence of interaction phenomena among the pedestrians, and of the pedestrians with the
structure. These two types of interaction can be investigated based on the theory of synchronisation of non-linear dynamic systems. The first occurs when a number of pedestrian walk very
close to each other, to the point of synchronising their walking frequency and phase, and it is
therefore related to some crowd density parameter. The second occurs when the structure vibrates with a sufficiently large amplitude, and at a frequency sufficiently close to the pedestrians
natural frequency, such to trigger their gait, and make them all walk at the structural vibration
frequency and all with the same phase. This second type of synchronisation is therefore related
to the dynamic characteristics of the footbridge. The possibility of occurrence of this two types
of synchronisation was illustrated in more detail by Pizzimenti and Ricciardelli (2005). However due to the complicated mathematics and to the difficulty in calibrating the parameters appearing in the models, an approach to the design of footbridges incorporating these two possible
synchronised types of behaviour is yet unavailable.
In fact, present design practice in most of the cases neglects the possible occurrence of interaction phenomena, and considers the pedestrians as an external load. Only in a few cases more
sophisticated loading patterns have been proposed and applied. Besides the possibility of establishing refined models incorporating possible synchronisation mechanisms, it is clear that the
use of simple loading models based on the tools of linear structural dynamics is appealing for
the designer, therefore it is of interest to assess their accuracy and to set the limits of their applicability.
In this paper two simple loading models will be presented, not including interaction phenomena. The first model is to be used for short span bridges, the second for longer ones. The results
of tests aimed at the calibration of the parameters required for the implementation of the two
models will be presented, together with the results of some other tests aimed at the definition of
the boundaries of applicability of linear loading models.

190

EVACES07

()

1.0

0.5

Eq. (2)
Eq. (3)
Numerical

0.0
0

Figure 1. Transient response reduction factor

2 LOADING MODELS
2.1 Short spans
For a short span footbridge the worst load condition proves to be the transient action of one pedestrian crossing the bridge. This because among the effects of additional pedestrians, the increase in
damping dominates on the increase in excitation. The additional pedestrians, in fact, cannot be
considered as external shakers, as with their movement they also provide out-of-phase forces, to be
seen as further damping for the system. This additional damping is minimised when all the pedestrians step in perfect phase with each other, but such condition is quite unusual when the pedestrians are free to chose their walking frequency and phase.
For this reason, many codes of practice (among which the 1995 version of Eurocode 5) provide for supported beams an equation for the maximum transient acceleration response to a
resonant sinusoidal load in the form:

amax M DLF
a M
=
() = max
()
W

W stat

(1)

in which M is the mass of the structure, W is the weight of the pedestrian, usually taken as 700
N, DLF is the Dynamic Load Factor and the structure damping ratio. The non-dimensional acceleration response is given in eq. (1) as the product of the stationary response and a transient
response reduction factor . This factor is a function of the structure damping and of the number of loading cycles n=L/l, L being the bridge span and l the wavelength of the load. In particular, the transient response reduction factor proves to be dependent only on the product of these
two quantities through the parameter =2n. The equation used in the codes of practice for ()
is derived form a paper of Blanchard and co-workers (1977), in which the transient response reduction factor was obtained by numerical integration of the equation of motion of the structure:
() = 1 e 2

(2)

More recently, an approximated closed form solution of the equation of motion was found
under the assumption of a sinusoidal mode shape, whose derivation is not shown here, bringing:

() =
1 + 2

+ atan
2

2
1+ + e

(3)

Eqs. (2) and (3) are compared in figure 1 in the practical range of the parameter, together
with the exact solution, coming from the numerical integration of the equation of motion. It appears that eq. (2) tends to overestimate the response, while eq. (3) proves quite accurate.
In addition to eq. (3), other equations were found for the approximate response away from
resonance, whose derivation is also not shown here. In the simplest form:

191

(amax x M )/(W x DLF )

Experimental Vibration Analysis for Civil Engineering Structures

100

Eq. (3)
Eq. (4)

Eq. (4)

50

0
0.0

0.5

1.0

1.5

2.0

Figure 2. Resonance curve for =0.01 and =1.2

amax M
2 2
= DLF
= DLF ()
W
1 2

(4)

where =fp/fo is the ratio of the frequency of excitation (walking frequency) to the natural frequency of the bridge.
Eqs. (3) and (4) allow plotting an approximated resonance curve for the transient response of
a supported bridge subjected to a sinusoidal moving load. An example of such a curve is given
in figure (2) for =0.01 and =1.2.
2.2 Long spans
For a medium to long span footbridge the worst load condition proves to be that of the stationary action of a crowd. For the particular case in which the density of the crowd is low enough
for the pedestrians not to interact with each other, therefore give rise to a synchronised behaviour, a loading model was presented by Ricciardelli (2005). Based on this, the RMS stationary
acceleration response of a supported bridge can be written as:

aRMS M
8 f o S Fe ( f o )
= BL

W
W

(5)

where foSFe( fo) is the non dimensional power spectrum of the equivalent load associated with
one pedestrian, is the density of pedestrians, and L and B are the span and width of the bridge.
The non dimensional power spectrum of the equivalent load associated with one pedestrian
is obtained as the integral of the power spectrum foSF( fo/fp) of the load associated with one pedestrian centred at the frequency fp weighted through the probability density p( fp) of the walking frequency:

f o S Fe ( f o ) = f o S F ( f o f p ) p ( f p ) df p
0

(6)

3 EXPERIMENTAL CALIBRATION OF THE MODEL PARAMETERS


The loading models presented in the previous section contain parameters that have to be calibrated experimentally. In particular the loading model of eq. (1), besides the knowledge of the
structural damping , requires knowledge of the DLF and of the load wavelength l, needed for
calculating the parameter . This is equal to the stride length for the vertical excitation, and
twice that for the horizontal excitation. On the other hand, the loading model of eq. (5), besides
the structural parameters B, L, fo, , requires knowledge of the pedestrian weight W and density
, of the load spectrum fSF( f), and of the probability density function of the walking frequency
pfp( fp).
Some of these parameters are available, others are not. In the following some experimental
results will be presented, and discussed together with those available in the literature.

192

EVACES07

Table 1. Statistical properties of the sample


n
Males
Females
TOTAL

62
54
116

W [N]
mean
RMS
757
106
552
88
646
129

mean = 1.835
STD = 0.172

H [cm]
mean
RMS
176
5.9
160
10.2
169
9.2
mean = 1.41
STD = 0.224

Age [yrs]
mean
RMS
28
13
26
12
27
12
mean = 0.768
STD = 0.098

1.15 1.35 1.55 1.75 1.95 2.15 2.35

0.75 0.95 1.15 1.35 1.55 1.75 1.95 2.15

0.52 0.58 0.66 0.74 0.82 0.90 0.98 1.06

f p [Hz]

v [m/s]

l [m]

Figure 3. Statistical distribution of the walking characteristics

3.1 Walking parameters


Data concerning the walking parameters (walking frequency fp, walking speed v and stride
length l) are available mainly from Biomechanics, with some additional results coming from research in Civil Engineering. It is clear that walking characteristics are strongly related to the
body characteristics and lifestyle. Therefore, tests performed in different locations, and in different conditions may give different results.
Some additional tests were recently carried out at the University of Reggio Calabria by Pansera (2006). These consisted in the measurement of the walking parameters of a sample of 123
individuals. Each of them was asked to walk naturally over a length of 72 m on a horizontal,
wide sidewalk. Walking time was measured and number of steps counted, and the average walking speed and stride length were calculated based on these. Out of the initial sample of 123, a final sample of 116 was selected, containing only the individuals who had shown a natural behaviour during the test. The statistical characteristics of the sample are given in table 1.
The influence of the characteristics (weight, height and age) of the walker on the walking parameters was first analysed, and it was found that there is a minor decrease of the walking frequency with weight (0.0002 Hz/N) and age (0.0012 Hz/yr), and a minor decrease of stride
length with age (0.03 cm/yr), whereas all the other parameters are sensibly independent of each
other. This information allows extrapolating the results provided below to populations with
characteristics different from those of table 1.
In figure 3 the histograms of walking frequency, walking speed and stride length are shown,
together with their Gaussian fits. In the histograms the walking frequency and stride length are
those associated with two successive steps, therefore they coincide with the vertical loading frequency and wavelength. For the horizontal loading, frequency halves, and the wavelength of the
load is twice the stride length. The mean and RMS values of the walking parameters are also
summarised in table 2 and compared with those of other authors. The Gaussian distribution fits
rather well the walking frequency data, and reasonably well the walking speed data, while it
seems inappropriate for the stride length. From table 2 it can be seen that the mean value of the
walking frequency generally ranges between 1.8 Hz and 2.0 Hz, and the RMS value between
0.11 Hz and 0.186 Hz. Less results are available for the stride length and walking speed, and in
addition to the study reported here, only that of Sahnaci and Kasperki (2005) gives separate data
for the two genders. The mean value of the stride length ranges between 0.71 m and 0.77 m,
which makes unjustified the value of 0.9 m adopted by Eurocode 5 and other codes of practice.
For some applications it is of interest the relationship between the three walking parameters
fp, l and v. Ingolfsson et al. (2007) suggested that a general law of the type y=axb be used to re-

193

Experimental Vibration Analysis for Civil Engineering Structures

Table 2. Statistics of the walking parameters


n
Males
Females
TOTAL
Matsumoto et al. (1978)
Wheeler (1982)
Bachmann & Ammann (1987)
Kerr & Bishop (2001)
Sahnaci & Kasperski (2005)
Zivanovic et al. (2005)
Pachi & Tianjian (2005)

62
54
116
505

40
251
1976
400

fp [Hz]
mean
RMS
1.812
0.172
1.860
0.168
1.835
0.172
1.99
0.173
2.0
2.0
0.18
1.9
1.82
0.12
1.87
0.186
1.8
0.11

l [m]
mean
0.796
0.734
0.768
0.75

0.75

1.5

0.07

1.37

0.71

0.15

1.3

2.0

2.0
v = 0.754f p + 0.024

URC
DTU
Terrier et al.

1.5
v [m/s]

1.5
v [m/s]

RMS
0.095
0.091
0.098

v [m/s]
mean
RMS
1.44
0.117
1.37
0.226
1.41
0.224

1.0
URC
DTU
Terrier et al.

0.5

1.0
0.5

0.0

v = 1.835l - 0.0007

0.0
0.0

0.5

1.0
1.5
f p [Hz]

2.0

2.5

Figure 4. Relationship between walking frequency


and walking speed

0.0

0.2

0.4

0.6
l [m]

0.8

1.0

1.2

Figure 5. Relationship between stride length and


walking speed

late any two of the three walking parameters. This as the advantage of allowing fulfilment of the
natural relationship between frequency, speed and stride v=fpl. On the other hand the tests discussed here revealed that walking speed can be reasonably assumed a linear function of both
walking frequency and stride length, while the latter parameters are quite uncorrelated with each
other. Such result is shown through figures 4, 5 and 6. In figure 4 the walking speed is shown as
a function of the walking frequency. In addition to the results of the tests carried out at the University of Reggio Calabria (indicated as URC), eight points are shown, coming from similar
tests carried out at the Danish Technical University (indicated as DTU), and other eight points
coming from the tests presented by Terrier et al. (2005). Together with the experimental points,
also the linear fit is shown, based only on the URC measurements. Such fit shows that the walking velocity is almost proportional to the walking frequency, therefore it seems reasonable to
neglect to constant value of 0.024 m/s and assume a relationship of the type v=0.754fp, where
the slope of the line is in good agreement with the measured mean stride length of 0.768 m. In
figure 4 the lines of equation v=[mean(l)RMS(l)]fp and v=[mean(l)2RMS(l)]fp are also
shown, which define the confidence ranges associated with the relationship between walking
speed and stride length.
A similar approach can be applied to analyse the relationship between stride length and
walking velocity (figure 5). Also in this case, when the constant value of -0.0007 m/s is neglected the linear relationship v=1.835l is obtained from the URC measurements, whose slope
is in perfect agreement with the measured mean walking frequency (see table 2). Also in figure
5 the confidence ranges of the relationship between stride length and walking velocity are
shown, through the lines of equation v=[mean(fp)RMS(fp)]l and v=[mean(fp)2RMS(fp)]l.

194

EVACES07

1.2
1.0
l [m/s]

0.8
0.6
0.4

URC
DTU
Terrier et al.

0.2
0.0
0.0

0.5

1.0
1.5
f p [Hz]

2.0

2.5

Figure 6. Relationship between walking frequency and stride length

Finally, in figure 6 the relationship between the walking frequency and the stride length is
shown. In this case the points are quite scattered, without a clear trend. In the figure a line corresponding to the mean stride length from the URC measurements, together with the constant values [mean(l)2RMS(l)] are also shown.
3.2 Walking forces
Depending on whether the loading model of eq. (1) or that of eq. (5) is used, a deterministic or a
probabilistic description of the walking forces is needed. The former is given through the DLFs
associated with the different harmonics of the vertical and horizontal load. The latter is given
through the power spectra of the vertical and lateral load appearing in eq. (6).
In table 3 a summary of the mean and characteristic values of the first five vertical and lateral
DLFs is reported. For the vertical DLFs the characteristic value is that associated with a probability of non exceedance of 75%, while for the horizontal DLFs the characteristic value is that
associated with a probability of non exceedance of 95%. The vertical DLFs are those reported
by Willford and Young (2005), who derived them from the analysis of measurements reported
by other authors. The horizontal DLFs are have been derived from the measurements of Pizzimenti (2005), also presented by Pizzimenti and Ricciardelli (2005). The first vertical DLF
shows a pronounced dependency on the walking frequency, which is reported in table 3. On the
other hand, the second, third and fourth vertical DLFs and the first and third horizontal DLFs
show only a moderate variability with frequency, associated with a large scatter of the data. In
table 3 both the variability with frequency and the suggested constant value are reported. The
remaining DLFs do non show any dependency on walking frequency.
The results of the measurements of Pizzimenti (2005) also revealed that the power spectrum
of the horizontal excitation around the n-th harmonic of the horizontal walking frequency can be
reasonably assumed to have a Gaussian shape, described through the equation:
f nf p 1 2
f S Fhn ( f ) = Cn W 2 exp 2

Bn

(7)

The parameters appearing in eq. (7) can be derived from the work of Pizzimenti and Ricciardelli (2005), and are shown in table 4. To the authors knowledge, no data are available for the
spectra of the vertical component of the load.
3.3 Crowd density
The applicability of the model of eq. (5) is bound to the hypothesis that crowd density is everywhere on the deck lower that a critical value, such that a larger density would cause interaction
among the pedestrians. No tests have yet been performed to characterise the critical density, and
it is suggested that a value between 0.2 and 0.3 pedestrians/m be used in the lack of other data.

195

Experimental Vibration Analysis for Civil Engineering Structures

Table 3. Dynamic Load Factors


n

Vertical

Horizontal

fp [Hz]

mean

75%

fp [Hz]

mean

95%

1.0-2.8

0.37fp-0.35<0.5

0.41fp-0.39<0.56

0.6-1.1

0.05fp -0.011
(0.028)

0.05fp+0.001
(0.041)

2.0-5.6

0.0051

0.0077

3.0-8.4

1.8-3.3

0.0086fp -0.0033
(0.017)

0.009fp+0.004
(0.023)

4.0-11

0.0056fp +0.069
(0.090)
0.0064fp +0.033
(0.069)
0.0065fp +0.013
(0.061)

1.1-2.2

0.0044fp+0.054
(0.070)
0.005fp+0.026
(0.054)
0.0051fp+0.01
(0.048)

2.4-4.4

0.0026

0.0044

>11

3.0-5.5

0.0077

0.011

Table 4. Parameters of the stochastic load model


1

Bn

0.0616

0.039

0.0288

0.037

0.035

Cn

0.0116

0.00086

0.0078

0.00064

0.00197

4 MEASUREMENTS ON A MODEL FOOTBRIDGE


For the loading models of eqs. (1) and (5) to be applicable, the amplitude of oscillation of the
footbridge has also to be small enough for the pedestrians not to be affected in their gait. To investigate the modifications occurring to the walking parameters due to the footbridge oscillation
tests were carried out at the Danish Technical University on a model footbridge, able to vibrate
in the vertical direction. These also allowed verifying the limits of applicability of eq. (1). The
footbridge, shown in figure 7, is a suspended span of 12 m, whose design data and final dynamic characteristics are reported by Ingolfsson (2006). For the purpose of matching the average walking frequency, an additional mass of 1075 Kg was added at midspan. There resulted a
first natural frequency of 1.95 Hz. The measured damping ratio proved to be dependent on the
amplitude of oscillation. In particular a value of 0.5% was measured for low amplitudes up to
3.5 cm, a value of 0.6% for intermediate amplitudes up to 5.5 cm, and a value of 0.8% for larger
amplitudes. In the following the value of 0.5% will be used, as in the tests the amplitude of oscillation never exceeded the value of 3.5 cm. The first modal mass in the configuration with the
additional mass at midspan was estimated to be in about 2020 Kg.
First, a set of tests was carried out to measure the undisturbed walking characteristics of the
eight test persons. These consisted in measuring the average walking frequency and speed and
stride length on a fixed floor, and the results are shown among those of figures 4 through 6. The
values obtained were considered as a reference for the measurements on the vibrating model
footbridge. Then three sets of metronome tests were carried out on the model footbridge. In test
#1 the metronome frequency was set equal to the bridge natural frequency, in test #2 it was set
equal to 90% that (1.75 Hz), and in test #3 equal to 110% (2.15 Hz). In the metronome tests the
persons were asked to try to keep the other walking parameters as natural as they could. Finally,
free walking tests were carried out, in which the persons were asked to walk on the bridge according to his/her own pace rate. Each of the tests above was repeated three times, and the results were averaged.
In table 5 the results of the free walking tests (on the stiff floor and on the vibrating footbridge) are compared. It can be noticed that there is a minor difference between the walking frequencies measured on the footbridge and those measured on the floor, to be considered as a random fluctuation. On the other hand, a systematic increase of the stride length is noticed when
the pedestrians walk on the footbridge. This ranges between 0 and 7%, and globally increases
with the amplitude of oscillation of the footbridge. In table 5 the measured response is also

196

EVACES07

Figure 7. Tests on the model footbridge

compared to that predicted through eqs. (1) and (5). Either the one or the other equation is used
in each case, depending on how close to each other were the walking and vibration frequencies.
The closed form solution of eq. (4) predicts the measured response with an error in the range of
34% to +42%. This is mainly explained through the fact that this equation, even though quite
accurate away from resonance, tends to become inaccurate as the system approaches resonance.
Other causes of inaccuracy are associated with the experimental setup, with the assumption of a
sinusoidal deformed shape, and with the short span of the footbridge. The closed form solution
of eq. (3), on the other hand, generally overestimates (between 42% and 77%) the measured response. This is mainly due to the fact that this equation applies in case of resonance, and in the
free walking tests this condition is generally not achieved.
In figure 8 the measured non-dimensional load detuning (difference between walking and
vibration frequencies) is plotted as a function of the theoretical non-dimensional load detuning
(that calculated based on the walking frequency on the stiff floor). Measurement of a zero detuning would indicate that the pedestrian has synchronised its gait to the footbridge vibration.
This situation, however, never occurred in the tests, suggesting that in none of the tests a
walker-structure synchronisation mechanism took place. The maximum measured acceleration
amplitude was 2.26 m/s2, corresponding to a displacement amplitude of 1.51 cm. It is concluded
that in this specific case of a short span, the synchronisation thresholds (both in terms of accelerations and displacements) are larger than the values above.
Finally in table 6 the results of the metronome tests are presented. First it is noted that there
is a discrepancy between the metronome frequency and the walking frequency, indicating that,
especially in the resonance tests the walkers couldnt fully match the triggered frequency. For
the resonance tests the measured acceleration is compared with that calculated through eq. (3).
The predicted acceleration always exceeds the measured one. For those walkers who could better match the metronome frequency (subjects # 2,4,5,7), the overestimation of the measured response through eq. (3) is in the range of 26% to 64%. In this case the overestimation of the response by eq. (3) is justified by the fact that when forced to follow the metronome frequency,
the walkers tend to modify their walking parameters (see the increase in stride length), therefore
modify the action to the footbridge. For the off-resonance tests the measured accelerations are
compared with those calculated through eq. (4). In this case the prediction error is larger when
the walking frequency is lower than the vibration frequency (up to 63%), and smaller when the
walking frequency is larger than the vibration frequency (as low as 4%).

197

Experimental Vibration Analysis for Civil Engineering Structures

Table 5. Free walking results


n

W
[N]

H
[m]

gender

559

1.60

647

fp [Hz]

a [m/s2]

l [m]

ratio

floor

bridge

floor

bridge

meas.

calculated

1.72

1.70

0.69

0.69

0.31

0.25 (eq. 4)

0.81

1.77

1.93

1.94

0.71

0.76

1.68

2.98 (eq. 3)

1.77

736

1.76

1.79

1.82

0.83

0.89

1.24

0.86 (eq.4)

0.69

785

1.80

1.88

1.83

0.85

0.90

1.98

2.81 (eq. 3)

1.42

697

1.74

1.87

1.90

0.81

0.87

1.95

2.76 (eq. 4)

1.42

687

1.80

1.82

1.82

0.83

0.85

1.48

2.52 (eq. 3)

1.70

736

1.78

1.95

1.90

0.85

0.89

2.26

2.86 (eq. 4)

1.27

1010

1.88

1.72

1.70

0.77

0.79

0.67

0.44 (eq. 4)

0.66

[( f p -f o)/f o]theor
-0.15

-0.1

-0.05

-0.05

-0.1

[(fp -f o)/f o]meas

-0.15

Figure 8. Measured vs. theoretical detuning

5 CONCLUSIONS
Two loading models for footbridges have been presented, to be used when no synchronisation
phenomena take place on the bridge, neither among the walkers, nor of the walkers with the
bridge oscillation. Values of the parameters contained in the models have been provided, either
as found in the literature, or as obtained by the authors experimentally. Finally, the results of
measurements on a flexible model footbridge have been presented, mainly meant at investigating the variation of the walking parameters due to the oscillation of the bridge. It was noticed
that when free to choose their pace rate, the walkers never modified it, with respect of what it
was on the fixed floor, at least for acceleration amplitudes up to 2.26 m/s2 and displacement
amplitudes up to 1.51 cm. On the other hand, they tend to increase the stride length as the amplitude of oscillation increases. This result is probably related to the short span of the bridge,
and it should not be excluded that on a longer span the same response amplitudes could cause
synchronisation.
Comparison of the measured and predicted response showed that the loading model proposed for short span bridges is generally on the safe side, but also that in particular cases it can
overestimate the response by more than 50%. This result is partly due to the inaccuracies associated with the experiments, partly due to the very short span of the model bridge. On the other
hand, for design purposes eq. (3) is meant to provide the maximum response the bridge can
reach when crossed by one pedestrian at the time. Therefore, its accuracy needs to be checked
based on the maximum of the responses measured for a large population of walkers. When this
is done, the inaccuracy tends to reduce and becomes acceptable for practical applications.

198

EVACES07

Table 6. Metronome walking results


fmetr=1.95 Hz

fmetr=2.15 Hz

fmetr=1.75 Hz

f
[Hz]

l
[m]

a
[m/s2]

Eq. (3)
[m/s2]

f
[Hz]

l
[m]

a
[m/s2]

Eq. (4)
[m/s2]

f
[Hz]

l
[m]

a
[m/s2]

Eq. (4)
[m/s2]

1.85

0.73

1.23

2.41

1.75

0.72

0.455

0.364

2.08

0.73

0.666

1.07

1.94

0.76

1.82

2.98

1.77

0.72

0.656

0.459

2.08

0.77

0.697

1.11

1.88

0.94

2.01

2.67

1.75

0.89

1.00

0.438

2.11

0.89

1.38

1.20

1.96

0.90

2.36

3.21

1.75

0.87

1.32

0.491

2.15

0.88

1.38

1.01

1.96

0.88

1.93

2.88

1.77

082

0.622

0.486

2.14

0.89

1.11

0.919

1.98

0.87

1.97

2.95

1.77

0.90

1.08

0.504

2.11

0.89

0.982

1.02

1.96

0.94

2.30

2.89

1.77

0.85

1.09

0.394

2.13

0.85

1.26

1.02

1.89

0.85

2.23

4.04

1.73

0.84

0.781

0.451

2.10

0.84

1.16

1.56

REFERENCES
Bachmann H, Ammann W. 1987, Vibrations in structures induced by man and machine, IABSE Structural Engineering Documents 3E
Blanchard J., Davies B. L., Smith J. W. 1977. Design criteria and analysis for dynamic loading of footbridges, Symposium on dynamic behaviour of bridges, Crowthorne, Berks, TRRL SR275
European Committee for Standardization (CEN) 1995. Eurocode 5, 1995: Design of timber structures.
ENV 1995-2 , Brussels
Ingolfsson E.T. 2006. Pedestrian-induced vibrations of line-like structures. Masters Thesis, BYD.DTU,
Technical University of Denmark
Ingolfsson E.T., Georgakis C.T., Jonsson J., Ricciardelli F. 2007. Vertical footbridge vibrations: towards
an improved and codifiable pedestrian loading model. Third International Conference on Structural
Engineering, Mechanics and Computation, Cape Town
Kerr S.C., Bishop N.W.M. 2001. Human induced loading on flexible staircases. Engineering Structures,
Elsevier, 23, 37-45
Matsumoto Y., Nishioka T., Shiojiri H., Matsuzaki K. 1978. Dynamic design of footbridges, IABSE Proceedings, Aug 1978, 17/78, 1-15
Pachi A., Ji T. 2005. Frequency and velocity of people walking. Structural Engineer, 83(3), 36-40
Pansera A. 2006. Analisi sperimentale delle caratteristiche del cammino ed azione dei pedoni sulle
passerelle pedonali. BSc Thesis, Univesrity of Reggio Calabria (in Italian)
Pizzimenti A.D. 2005. Analisi sperimentale dei meccanismi di eccitazione laterale delle passerelle ad
opera dei pedoni. Ph.D. Thesis, University of Catania (in Italian)
Pizzimenti A.D., Ricciardelli F. 2005. Experimental evaluation of the dynamic lateral loading of footbridges by walking pedestrians. 6th European Conference on Structural Dynamics, Paris
Ricciardelli F. 2005. Lateral loading of footbridges by walkers. Footbridge 2005 2nd International
Conference, Venice
Sahnaci C., Kasperski M. 2005. Random loads induced by walking. 6th European Conference on Structural Dynamics, Paris
Terrier P., Turner V., Schutz Y. 200p. GPS analysis of human locomotion: further evidence of long-range
correlations in stride-to-stride fluctuations of gait parameters, Human Movement Science, 24(1), 583591
Willford M.R., Young P. 2005. Improved methodologies for prediction of footfall-induced vibration. 6th
European Conference on Structural Dynamics, Paris
Wheeler J. E., 1982. Prediction and control of pedestrian-induced vibration in footbridges, J. Struct.Eng.
ASCE, 108 (9), 2045-2065
Zivanovic S., PAvic A., Reynolds P., Vuiovic P. 2005. Dynamic analysis of lively footbridges under everyday pedestrian traffic. 6th European Conference on Structural Dynamics, Paris