|Software for Operational
Damage Detection of the Dogna Bridge in Italy
This case study presents a novel application of the new
Damage Detection plugin available in
ARTeMIS Modal Pro. It
shows how the plugin is used with measurements obtained from the Dogna
bridge in Italy in the spring of 2008 by Professor Morassi of the
Dipartimento di Ingegneria Civile e Architettura at the Università degli
Studi di Udine, Italy, and Professor Benedettini of the Dipartimento di Ingegneria Civile, Edile-Architettura e Ambientale
Università degli studi dell'Aquila,
Italy. Structural Vibration Solutions A/S gratefully acknowledges both
professors for their help with this
case study and for sharing the measurements, photos and relevant information
about this bridge.
Dogna Bridge and the Exceptional Flooding of 2003
The Dogna bridge crosses the River Fella and connects the villages of
Crivera and Valdogna (Dogna) in Friuli Venezia Giulia, a region located
in the North East of Italy.
Figure 1: The Dogna bridge crossing the River Fella.
As shown in Figure 1, the bridge is a four-span, single-lane concrete bridge. The length of each
span is 16 m, and the lane is about 4 m width. The bridge deck is made of a
reinforced concrete (RC) slab 0.18 m thick, supported by three
longitudinal RC beams with a rectangular cross-section 0.35 x 1.20 m. The beams
are simply supported and rest at the ends on thin metallic sheets.They are connected
at the supports, at midspan and at span quarters with rectangular cross-section
RC diaphragms (0.3 x 0.7 m). Each pier is a RC wall
about 1.5 m thick, 4.5 m deep, and around 3.6 m high.
The abutments consist of RC walls, 1 m thick. Piers and abutments were
built on cast-in-place concrete piles of 1 m in diameter and 18 m in length.
On August 31, 2003 the bridge suffered severely due to an exceptional flood
of the Fella River. At that time, due to the material deposited upstream,
the deck structures of the bridge were overtop by the flow of the water as
seen in the pictures below.
Figure 2: Pictures of the actual flooding on August 31, 2003.
A visual inspection conducted on the bridge revealed no apparent
deterioration of the slab and beams, whereas a state of degradation was
noticed on support bearing side pier. For traffic safety reasons, the Dogna
was demolished on May 2008 and has been replaced by a new bridge built about
200 m downstream.
The Experimental Campaign
A testing campaign was carried out from April 2 to April 11, 2008 and
consisted of a series of tests of progressive damage of one of the bridge
spans. The tested span was made independent of the adjacent span by removing
the deck-joint of the pier. Moreover, the asphalt overlay of
about 0.1 m thickness was also removed before testing. All the tests were
carried out under similar environmental and weather conditions so that the
influence of temperature and humidity on the dynamic modal parameters would
not be a factor of significant importance. Figure 3 shows the bridge span during the tests and
how the progressive damage was introduced artificially.
Figure 3: Asphalt overlay was removed, and bridge span was made independent
of the adjacent span. Damage was artificially introduced.
Up to three
cuts of one of the supporting beams as well as the introduction of a damage
in the centerline of the side beam.
In this case study, three out of the seven damage cases have been considered. These
cases considered are shown in the table below, and the data was obtained
using an Ambient Vibration Testing acquisition system.
||April 3, 2008
before any artificial damage was introduced.
||April 3, 2008
after the first half cut of one of the beams, see Figure 3.
||April 4, 2008
after the first cut of the beam was completed.
||April 24, 2008
after three cuts of a beam and the introduction of damage
in the centerline of the side beam.
Table 1: List of the measurement campaign. One reference case and three
damaged are used in this case study.
Before the progressive damage was introduced artificially, a reference
Ambient Vibration Test was conducted for 20 minutes in 10 locations on the
bridge span. The total number of samples acquired was 475341. The figure 4
below shows the first four modes of the structure.
Figure 4: First 4 fundamental modes of the bridge span.
In order to establish the thresholds for being inside the safe zone, the
damage detection plugin needs multiple files that ultimately result in
multiple indicator values. In order to accomplish this, the undamaged
measurement record was divided into eight separate measurement files. The six first
were then used to create the reference state model, and the remaining two were
used to validate the performance of the reference state model.
Procedure for Establishing the Reference State Model
The following describes the steps required to establish the reference state
Define a new External Storage ARTeMIS Modal
project using e.g. the interactive setup tools or from a SVS
Configuration File. Here the configuration file used previously for the
above Operational Modal Analysis was re-used by replacing the original
undamaged measurement file with the first of the eight subdivided
In ARTeMIS Modal Pro, measurements and
associated results are stored as Analysis Sessions. This allows
the storage and presentation of historical development of the results
obtained from different measurements. Each time a measurement is
uploaded, ARTeMIS Modal Pro will create a new Analysis Session and
associate it to the uploaded measurement. The Signal Processing dialog in the Prepare
Data task is used to configure how the uploaded measurements for this
first Analysis Session should be processed. Here we chose to decimate
the measurements and end up selecting this initial Analysis Session as
the Master Session. The parameters in the Master Session are used for all other Analysis Sessions
that will be created in the project.
The remaining seven reference measurement files
are then uploaded. By switching to the Analysis History task, the uploading
of the files can be done simply by dragging them from the
Windows Explorer to the Analysis History window. The files are then
automatically uploaded and processed. In Figure 5 below, the top left
picture shows the list of reference Analysis Sessions. The green check
mark indicates that the signal processing is complete for each session.
When all the reference measurements are
uploaded and processed, it is possible to configure how the damage
detection algorithms should estimate the reference state model. In Figure
5, the Damage Detection
configuration dialog is shown. In this dialog, the Start and End Reference
Analysis Sessions are selected as the first six out of the eight, and
the Start button is pressed. This will create the reference state model,
the statistical thresholds and damage indicators of the eight reference
measurements, see Figure 6.
Figure 5: Establishing the reference state model. It
consists of the six
measurements uploaded and stored in their own Analysis
In the Analysis History task window, the Damage Detection
configuration dialog is used to select the range of Analysis Sessions
measurements from in the reference state model.
Once established, the Chi
Square damage indicators of each of the
six reference measurements are calculated
and used to estimate the
Safe-Zone and Critical-Zone thresholds.
Figure 6: Once the reference state model and
thresholds are estimated, all uploaded measurements are presented by the
Here the six first inside the gray zone are the ones used to establish the
reference state model, and the remaining two are used to validate
that new data from the same reference state is actually within the safe zone
as well. In addition, it is possible to investigate the null space
estimation of the reference state model. In the diagram called Reference
State Validation, singular values of the null space are presented as
the upper dark-gray zone. The singular values should preferably be zero in
this region, which is the case here, where no yellow singular values
can be seen above the indicator of the selected system order.
Once the Reference State Model has been estimated
along with the thresholds, it is a simple matter to test the three damage
cases presented in Table 1. The measurement files for each of the damage
cases are simply dragged from the Windows Explorer to the Analysis History
window that will process them automatically and present the associated
damage indicators. If there is a significant change of the dynamics, the
indicators will pass the thresholds. Changes in the dynamics of the
structure beyond the threshold levels will be shown as red colored
bars on the chart. Larger bars indicate very significant changes on the
structural system which can be related to damage of the structure. In Figure 7 below, the indicators are shown for the three damage cases, and
they all indicate a significant change of the dynamic behavior of the
structure. Also, it is clear that the damage is increasing from the first to
the last damage case, as expected.
Figure 7: All damage indicators presented. The first
8 green ones are from the reference (undamaged) measurements and the last
ones are from the three cases of progressive damage, D1, D2 and D7.
This case study demonstrates the use of the
Detection plugin in ARTeMIS
Modal Pro. A reference state model is constructed from uploading a
series of measurement files all describing the reference state of the
structure. Once the reference state model is constructed, damage detection
is performed simply by dragging measurement files, acquired from the
potentially damaged structure, into the Analysis History window. The damage
indicators are then automatically calculated and presented either in green
(safe state), yellow (critical state) or red (unsafe state).
M. Döhler and F. Hille:
Subspace-based damage detection on steel frame structure under
Proc. 32th International Modal Analysis Conference
(IMAC) Orlando, Florida USA, 2014.
F. Benedettini, A. Morassi: Dynamic testing, structural identification
and damage detection on Dogna’s bridge. Proceedings of the 8th
International Conference on Structural Dynamics, EURODYN 2011, Leuven,
M. Dilena, A. Morassi , M. Perin: Dynamic
identification of a reinforced concrete damaged bridge. Mechanical
Systems and Signal Processing, 2011.
M. Dilena, A. Morassi: Dynamic testing of a
damaged bridge. Mechanical Systems and Signal Processing, 2011.
M. Dilena, M.P. Limongelli, A. Morassi: Damage
localization in bridges via FRF interpolation methods. Mechanical
Systems and Signal Processing, 2014. (Under revision)
M. Döhler, L. Mevel, and F. Hille: Subspace-based damage detection under
changes in the ambient excitation statistics. Proc. 18th IFAC World
Congress, Milan, Italy, 2011.
F. Hille, M. Döhler, L. Mevel and W. Rücker: Subspace based damage
detection methods on a prestressed concrete bridge. Proc. 8th Int.
Conf. on Structural Dynamics EURODYN, Belgium, 2011