Experimental Vibration Analysis for Civil Engineering Structures

413

Modal identification of the Parque da Paz roadway and railway

viaducts
J. Rodrigues & M. Ledesma
LNEC Structures Department, Lisbon, Portugal

ABSTRACT: This paper describes and presents the results of modal identification tests performed in two viaducts, located in Parque da Paz in Almada, in the South part of the Lisbon
metropolitan area. One of them is a roadway structure and the other is a railway viaduct for the
light train of the new public transportation system that serves the city of Almada and its vicinity.
Both viaducts have equal spans and similar structural systems. The dynamic tests consisted in
ambient vibration measurements and were performed as part of the reception tests carried out
before the opening of the viaducts to the traffic, which also included static load tests. The ambient vibration tests provided a considerable amount of acceleration data which was analyzed using an output-only modal identification method to identify the modal properties of both structures. The dynamic characteristics evaluated for both viaducts are compared with each other and
with the values computed with finite element models.
1 INTRODUCTION
Modal identification of civil engineering structures has a great interest for the calibration of analytical and numerical models that are to be used in studying the behavior of those structures to
the different types of dynamic loads they have to withstand, like traffic, wind or seismic loads.
Structural health monitoring is also an area where modal identification of civil engineering
structures is having an increasing interest. Since the dynamic characteristics of a structure, like
frequencies, damping and mode shapes, are directly related to the stiffness, mass and energy
dissipation mechanisms and to their spatial distribution, they reflect the actual global condition
of a structural system (structural health). Situations of structural damage can, eventually, be detected from changes in the identified dynamic properties.
The identification of the modal properties of civil engineering structures involves the analysis
of experimental data obtained in dynamics tests. Those tests can be performed using several experimental techniques, from forced vibration tests, where different types of excitation equipments can be used, to ambient vibration tests, which rely on the natural sources of excitation
like wind or traffic, Paultre et al. (1995), Felber & Cantieni (1996).
To apply modal identification in structural health monitoring it is important that the adopted
techniques are able to evaluate the dynamic characteristics without the need to interrupt the normal use of a structure (for instance in a bridge, without interruption of the usual traffic). With
natural input testing or ambient vibration testing it is possible to satisfy this requirement, since it
consists in the measurement of the dynamic responses induced by the loads to which the structures are usually subjected, like wind or traffic. Therefore, from the various types of dynamic
tests that can be applied in civil engineering structures, ambient vibration testing is the more
reasonable approach if there is the purpose to apply it periodically, or even continuously with
permanent equipment, during the lifetime of a structure.
Today there is a considerable development at the level of the equipments for measurement of
the dynamic response of structures. Many sensors have a high sensitivity and data acquisition

414

EVACES07

systems a finer resolution, both allowing to measure adequately structural responses with very
small amplitudes.
There are also several modal identification techniques specifically developed for output-only
situations, Andersen et al. (1999), Peeters (2000). Some of those methods require a considerable
computational effort, but which is easily handled with the computing speed and data storage capacity of todays computers.
At the Structures Department of LNEC, there has been an increasing activity in the application of ambient vibration tests and output-only modal identification, for evaluation of the dynamic characteristics of bridges. In recently built bridges, for which LNEC is usually required to
do reception load tests, there has been a major concern in doing an evaluation of the dynamic
properties as complete as possible, that can be considered as corresponding to a reference situation of the structure, to be taken into account for monitoring its condition during its lifetime.
The work presented in this paper was developed within the reception load tests performed in
two viaducts, located in Parque da Paz in Almada, in the South part of the Lisbon metropolitan
area. One of them is a roadway structure and the other is a railway viaduct for the light train of
the new public transportation system that serves the city of Almada and its vicinity. Those tests
were performed before the viaducts were opened to the traffic and included static load tests and
dynamic tests, Rodrigues and Ledesma (2006, 2007).
In this paper, after this introduction, the Parque da Paz roadway and railway viaducts are
briefly described. Next, the equipment used in the tests and the testing procedures are presented.
The following items address the modal identification methods that were used to process the recorded accelerations and present the main results that were obtained. Finally the characteristics
estimated from the tests in the two viaducts are compared with each other and with the ones
evaluated with finite element models and some conclusions are made about the work that was
developed.
2 DESCRIPTION OF THE VIADUCTS
The Parque da Paz roadway and railway viaducts are parallel to each other, having both a total
length of 170 m, between the axes of the supports at the abutments, divided in 7 spans with the
following sequence of lengths: 21 m; 24 m; 24 m; 32 m; 24 m; 24 m and 21 m. The roadway
viaduct is located on the North side and the railway viaduct on the South side. The deck of the
two viaducts is continuous in reinforced concrete with longitudinal pre-stress.
The roadway viaduct deck has a cross section with a width of 9.5 m, composed by two beams
connected by an intermediate slab with a thickness of 0.3 m. This slab extends laterally, beyond
the beams, in cantilevers with a thickness of 0.2 m at their edges. The longitudinal beams have a
trapezoidal cross section with a height of 1.6 m and a width that, at the middle of the spans, varies from 0.8 m at the base and 1.0 m at the top slab, which is gradually increased to, at the sections over the piers, 1.3 m at the base and 1.5 m at the top slab.
The railway viaduct deck has a cross section which is very similar to the one of the roadway
viaduct. It is only somewhat wider, going up to a width of 10.2 m.
The piers of both viaducts have a cross section which is approximately rectangular with 1.0 m
x 3.0 m. The height of the piers varies from 7.0 m to 9.0 m. On the top of each pier the width of
the cross section gradually increases from 3.0 m up to about 7.4 m.
The deck is supported with elastomeric bearings at the abutments and at the extreme piers P2
and P7. At the other piers, the connection between the deck and the piers is continuous.
The foundations of the piers and abutments are pile foundations.
An elevation view of the Parque da Paz viaducts is presented in Figure 1 and the crosssections are presented in Figure 2.

Figure 1. Elevation view of the Parque da Paz viaducts.

Experimental Vibration Analysis for Civil Engineering Structures

415

Figure 2. Cross sections of the Parque da Paz viaducts.

The structural design of the two viaducts was performed by the company GEG (2004). In
terms of the loads considered in the design of both viaducts, it is important to mention that the
roadway viaduct was designed for the loads specified in the Portuguese Structural Design Code
RSA (1983), while the railway viaduct was designed for a load consisting in trains with 8 axles
of 100 kN each, as specified by the owner of the structure.
A general view from the South side of the Parque da Paz viaducts is presented in Figure 3.
Figure 4 shows two photos of the viaducts, one of the upper part of the deck and the other of the
lower part of the deck, in both photos the roadway viaduct is on the left side and the railway
viaduct is on the right side.

Figure 3. General view from the South side of the Parque da Paz viaducts.

Figure 4. Views of the Parque da Paz viaducts from above and bellow the deck.

3 DYNAMIC TESTS
3.1

Equipment used in the tests

The dynamic tests of the Parque da Paz viaducts were performed with the following equipment:
- 15 Kinemetrics Uniaxial Episensor (ES-U) accelerometers;
- five power supply and signal conditioning units with respective cables for power and signal transmission;
- data acquisition equipment from Gantner Instruments, including an e.gate DP module and
4 e.bloxx A1-4 modules with 19 bits A/D conversion;
- a temperature sensor was used to measure the ambient temperature while the accelerations in the structure were measured;
- one laptop computer.

416

EVACES07

The ES-U accelerometers are force balance acceleration sensors from Kinemetrics. They
have a high dynamic range (greater than 145 dB) and low noise performance. Their bandwidth
goes from DC to 200 Hz, which is an important feature for dynamic testing of civil engineering
structures, especially in the case of long span bridges where the first vibration modes have low
frequencies. The ES-U accelerometers also have a user selectable full-scale range, which allows
using them in several different situations.
In order to use the ES-U accelerometers, a system for power supply and signal conditioning
was developed at the Centre for Scientific Instrumentation (CIC) of LNEC. This system comprises 5 power supply and signal conditioning units. Each of these units can be connected to
three accelerometers and contains two 12V batteries, analog filters and amplifiers so that the
user can select gain factors of 1, 2, 5, 50, 100, 200, 250, 400 and 1000. The units are connected
to the sensors with relatively short cables (20 m) and to the data acquisition system with long
cables that go from 50 m up to 200 m.
A scheme of all the equipment used in the dynamic tests of the Parque da Paz viaducts is presented in Figure 5.
accelerometers

power supply and signal

conditioning units

ES-U
ES-U
ES-U

unit 1

ES-U
ES-U
ES-U

unit 2

ES-U
ES-U
ES-U

unit 3

ES-U
ES-U
ES-U

unit 4

ES-U
ES-U
ES-U

unit 5
20 m

50 m
100 m

laptop computer
e.gate DP and
4 e.bloxx A1-4
ethernet

110 m
200 m
50 m

temperature sensor

Figure 5. Equipment used in the dynamic tests

In the dynamic tests of the Parque da Paz viaducts, the ES-U accelerometers were configured
with a sensitivity of 10 Volt/g and the gain factor at the power supply and signal conditioning
units was configured to 50. With this configuration, the minimum acceleration amplitude that
could be measured was 0.076 g.
3.2 Testing procedure
The testing procedure adopted in the ambient vibration measurements was similar in both viaducts with just some slight differences. In the roadway viaduct, accelerations were measured in
57 points of the deck and in the railway viaduct 56 points were instrumented.
The tests were performed in a total of 5 set-ups. The first 4 set-ups were performed to identify
the vertical and torsion modes of the viaduct, while the 5th sep-up was made to identify the
transverse modes. In the first 4 set-ups, from the 15 available sensors, 3 were considered as reference sensors and remained at the same points during all those 4 set-ups, while the remaining
12 roving sensors were placed in the other points, changing their position from set-up to set-up.
For the preparation of the tests, namely for the selection of the reference points, it was important to do some pre-test analysis with finite element models of the viaducts.
In each set-up of the ambient vibration tests, accelerations were measured during a total time
of 20 minutes. The records were acquired with a sampling frequency of 1000 Hz and were later
pre-processed with high-pass filtering at 0.5 Hz, low-pass filtering at 20 Hz, using an 8 poles
Butterworth filter, and decimation to a sampling frequency of 50 Hz.
The dynamic tests of the roadway viaduct were performed in October of 2006, before the viaduct was opened to the traffic; therefore, most of the vibrations measured during those tests
were due to wind and also to the road traffic in the vicinity of the viaduct. The tests of the railway viaduct were performed in March 2007; at that time there wasn't yet traffic in the railway
viaduct but the roadway one was already being used.

417

Experimental Vibration Analysis for Civil Engineering Structures

4 MODAL IDENTIFICATION
The ambient vibration tests data was analyzed, for modal identification purposes, using a frequency domain output-only modal identification method, the enhanced frequency domain decomposition method (EFDD), Brincker et al. (2001), implemented in the software Artemis Extractor, SVS (2002). The EFDD method is an improvement of the frequency domain decomposition method (FDD), Brincker et al. (2000).
The basic data for the FDD and EFDD methods is the matrix of spectral density functions of
the response of a structural system. These functions are usually estimated with the following
procedure: division of the response records in several, eventually overlapped, segments, whose
size determines the frequency resolution of the spectral estimates; application of a signal processing window to reduce the effects of leakage; computation of the DFT of the windowed data
segments trough the use of the FFT algorithm; computation of averaged auto and cross spectra
considering the DFTs of the data segments.
In the FDD method the spectral density functions matrix is, at each discrete frequency, decomposed in singular values and vectors using the SVD algorithm. By doing so, the spectral
densities are decomposed in the contributions of the different modes of a system that, at each
frequency, contribute to its response. In each frequency, the dominant mode shows up at the 1st
singular value spectrum and the other modes at the other singular values spectra. From the
analysis of the singular values spectra it is therefore possible to identify the auto power spectral
density functions corresponding to each mode of a system, which may include parts of several
singular values spectra, depending on which mode is dominant at each frequency. In the FDD
method, the mode shapes are estimated as the singular vectors at the peak of each auto power
spectral density function corresponding to each mode.
The EFDD method is similar to the FDD technique, with only some additional procedures to
evaluate the damping and to get enhanced estimates of the frequencies and mode shapes. In the
EFDD method the selection of the auto-spectra corresponding to each mode of a system is based
on the values of the MAC coefficient between the singular vectors at the resonance peaks and at
their neighboring frequency lines. Those SDOF auto-spectral density functions are then transformed back into the time domain by inverse FFT, resulting in auto-correlation functions for
each mode of a system. Enhanced estimates of the frequencies of the modes of a system are obtained from the zero crossing times of those auto-correlation functions. The damping coefficients are estimated from the logarithmic decrement of those auto-correlation functions. Finally,
the estimate of the mode shapes is also enhanced, considering all the singular vectors within
each SDOF auto-spectral density function, weighted with the corresponding singular values.
To apply the EFDD method to the data obtained in the Parque da Paz viaducts, the spectral
density functions of the accelerations were estimated with the FFT algorithm applied to windowed and overlapped samples with 2048 values. Since the sampling frequency of the records is
50 Hz (after pre-processing) the frequency resolution of the estimated spectra is 0.024 Hz.
Figures 6 and 7 show the averaged spectra of the first two singular values of the spectral density functions matrices of the, respectively, transverse and vertical accelerations measured in the
two viaducts.
The frequencies identified with the EFDD method for the vibration modes of the two viaducts
are also shown in Figures 6 and 7. In both cases a total of 10 vibrations modes were identified (6
transverse, 3 vertical and 1 torsion mode).

-30
-40

-50

-50
frequency (Hz)

-60

14.02 Hz

11.66 Hz

8.28 Hz

-20

-40

-70

5.05 Hz

0
-10

-30

-60

railway viaduct

amplitude (dB)
3.03 Hz

8.43 Hz

14.03 Hz

-20

4.97 Hz

0
-10

11.08 Hz

10

4.02 Hz

roadway viaduct

amplitude (dB)
3.38 Hz
4.05 Hz

10

frequency (Hz)

-70
0

10

11

12 13

14 15

Figure 6. Singular values spectra of the transverse records

10 11

12 13

14

15

418

EVACES07

-20
-30

-40

-40

-50

-50

-60

frequency (Hz)

7.75 Hz

3.03 Hz

6.67 Hz

-10

-30

railway viaduct

amplitude (dB)

7.45 Hz

-20

5.41 Hz

3.38 Hz

0
-10

4.24 Hz

10

6.13 Hz
6.87 Hz

amplitude (dB)

4.52 Hz

roadway viaduct
10

-60

frequency (Hz)

-70

-70
0

10 11

12 13 14 15

10 11 12

13 14 15

Figure 7. Singular values spectra of the vertical records

Figures 8 to 13 show the frequencies, damping ratios and shapes of the first six vibration
modes identified with the EFDD method for the Parque da Paz roadway and railway viaducts.
roadway viaduct (f = 3.38 Hz, = 1.5 %)

railway viaduct (f = 3.03 Hz, = 3.0 %)

Figure 8. Identified shapes of the 1st transverse mode (plan view)

roadway viaduct (f = 4.05 Hz, = 1.2 %)

railway viaduct (f = 4.02 Hz, = 1.5 %)

Figure 9. Identified shapes of the 2nd transverse mode (plan view)

roadway viaduct (f = 4.24 Hz, = 2.6 %)

railway viaduct (f = 4.52 Hz, = 1.0 %)

Figure 10. Identified shapes of the 1st vertical mode (elevation view)

Experimental Vibration Analysis for Civil Engineering Structures

419

roadway viaduct (f = 4.97 Hz, = 3.0 %)

railway viaduct (f = 5.05 Hz, = 2.9 %)

Figure 11. Identified shapes of the 3rd transverse mode (plan view)
roadway viaduct (f = 5.41 Hz, = 1.3 %)

railway viaduct (f = 6.13 Hz, = 0.6 %)

Figure 12. Identified shapes of the 2nd vertical mode (elevation view)
roadway viaduct (f = 6.67 Hz, = 1.3 %)

railway viaduct (f = 6.87 Hz, = 0.8 %)

Figure 13. Identified shapes of the 1st torsion mode (elevation view)

5 COMPARISON WITH FE MODELS

Finite element models of the Parque da Paz roadway and railway viaducts were developed using
the software SAP2000, CSI (2002). They are three dimensional, linear, elastic models of the
structures. The deck of the viaducts was modeled with shell and beam elements and the piers
were modeled with beam elements.
A general view of the developed finite element models is presented in Figure 14.
A comparison of the identified natural frequencies with the computed frequencies is presented in Figure 15. One can see that there is a good agreement between the identified and computed values, maybe with a larger discrepancy in the higher order transverse mode.

420

EVACES07

Figures 16 to 19 show the computed frequencies and shapes corresponding to the vibration
modes that were shown previously in Figures 8 to 13.
roadway viaduct

railway viaduct

Figure 14. Finite element models of the viaducts

roadway viaduct

16

12

transverse modes
vertical modes
torsion modes

14
computed frequency (Hz)

computed frequency (Hz)

14

railway viaduct

16

transverse modes
vertical modes
torsion modes

10
8
6
4

12
10
8
6
4
2

0
0

6
8
10
identified frequency (Hz)

12

14

16

Figure 15. Identified and finite element model frequencies

roadway viaduct (f = 3.41 Hz)

railway viaduct (f = 3.23 Hz)

Figure 16. Computed shapes of the 1st transverse mode (plan view)
roadway viaduct (f = 4.23 Hz)

railway viaduct (f = 3.95 Hz)

Figure 17. Computed shapes of the 2nd transverse mode (plan view)

6
8
10
identified frequency (Hz)

12

14

16

Experimental Vibration Analysis for Civil Engineering Structures

roadway viaduct (f = 3.96 Hz)

railway viaduct (f = 4.15 Hz)

Figure 18. Computed shapes of the 1st vertical mode (elevation view)

roadway viaduct (f = 5.09 Hz)

railway viaduct (f = 4.97 Hz)

Figure 19. Computed shapes of the 3rd transverse mode (plan view)

roadway viaduct (f = 5.57 Hz)

railway viaduct (f = 5.85 Hz)

Figure 20. Computed shapes of the 2nd vertical mode (elevation view)

roadway viaduct (f = 6.18 Hz)

railway viaduct (f = 6.13 Hz)

Figure 21. Computed shapes of the 1st torsion mode (elevation view)

421

422

EVACES07

6 CONCLUSIONS
The dynamic tests of the Parque da Paz roadway and railway viaducts, presented in this paper,
were conducted by LNEC before they were opened to the traffic, in order to evaluate their dynamic characteristics.
The ambient vibration tests and the application of an appropriate output-only modal identification technique were successful, since they allowed to identify several natural vibration modes
of the two structures.
As expected, the identified natural frequencies are similar in both structures, with just some
slight differences. The experimental results showed also a good agreement with the values computed with finite element models of the viaducts.
The identified dynamic characteristics of the Parque da Paz roadway and railway viaducts are
an important contribution to characterize their actual condition before they were opened to the
traffic. It is important to note that the natural excitation dynamic tests, which were performed
for modal identification purposes, can be carried out during the life-time of the structures without the need to impose traffic restrictions.
7 ACKNOWLEDGEMENTS
The authors would like to acknowledge the cooperation of Eng. Jos Diniz and Eng. Pedro
Coelho from Metroligeiro and of all the LNEC technicians that participated in the dynamic tests
of the Parque da Paz roadway and railway viaducts.
8 REFERENCES
Andersen, P., Brincker, R., Peeters, B., De Roeck, G., Hermans, L. & Kramer, C. 1999. Comparison of
system identification methods using ambient bridge test data. IMAC XVII, Kissimmee, USA, February
8-11.
Brincker, R., Zhang, L. and Andersen, P. 2000. Modal Identification from Ambient Responses Using
Frequency Domain Decomposition. IMAC XVIII. San Antonio, USA, February 7-10.
Brincker, R., Ventura, C. and Andersen, P. 2001. Damping Estimation by Frequency Domain Decomposition. IMAC XIX. Kissimmee, USA, February 5-8.
CSI. 2002. SAP2000 Integrated Finite Element Analysis and Design of Structures. Computers & Structures, Inc., USA.
Felber, A. & Cantieni, R. 1996. Advances in ambient vibration testing: Ganter bridge, Switzerland. Structural Engineering International, Vol. 3/96, pp. 187-190.
GEG - Gabinete de Estruturas e Geotecnia, Lda., 2004. Parque da Paz Viaduct, Design Documents and
Drawings (in Portuguese).
Paultre, P., Proulx, J. & Talbot, M. 1995. Dynamic testing procedures for highway bridges using traffic
loads, Journal of Structural Engineering, Vol. 121, N. 2, ASCE, pp. 362-376.
Peeters, B. 2000. System identification and damage detection in civil engineering. PhD thesis, KU Leuven, Belgium.
Rodrigues, J. & Ledesma, M. 2006. Load tests of the bridges of the Tagus river South bank light subway Static and dynamic tests of the Parque da Paz roadway viaduct. LNEC Report (in Portuguese).
Rodrigues, J. & Ledesma, M. 2007. Load tests of the bridges of the Tagus river South bank light subway Static and dynamic tests of the Parque da Paz railway viaduct. LNEC Report (in Portuguese).
RSA. 1983, Regulamento de Segurana e Aces para Estruturas de Edifcios e Pontes, Decreto-lei
n235/83, INCM, Lisboa (in Portuguese).
SVS Structural Vibration Solutions. 2002. ARTeMIS Extractor Handy. Release 3.1, Denmark.