Journal of Physics: Conference
Series
OPEN ACCESS You may also like
- Exploring vibration control strategies for a
Structural Finite Element Model Updating Using footbridge with time-varying modal
parameters
Vibration Tests and Modal Analysis for NPL Jose M. Soria, Ivan M. Díaz, Emiliano
Pereira et al.
footbridge – SHM demonstrator - Viscoelastic damper for vibration mitigation
of footbridges
Abbas Ali Alhasa, Mohammadreza Vafaei,
To cite this article: E Barton et al 2011 J. Phys.: Conf. Ser. 305 012105 Sophia C. Ali et al.
- Load Testing of GFRP Composite U-
Shape Footbridge
ukasz Pyrzowski, Mikoaj Mikiewicz, Jacek
Chrócielewski et al.
View the article online for updates and enhancements.
This content was downloaded from IP address 169.159.129.118 on 14/03/2024 at 23:35
9th International Conference on Damage Assessment of Structures (DAMAS 2011) IOP Publishing
Journal of Physics: Conference Series 305 (2011) 012105 doi:10.1088/1742-6596/305/1/012105
Structural Finite Element Model Updating Using Vibration
Tests and Modal Analysis for NPL footbridge – SHM
demonstrator
E Barton1, C Middleton2, K Koo2, L Crocker1, J Brownjohn2
1
Structural health monitoring, National Physical Laboratory, Hampton Road,
Teddington, Middlesex, TW11 0LW
2
The University of Sheffield, Department of Civil and Structural Engineering,
Vibration Engineering Research Section, Sir Frederick Mappin Building
Mappin Street, Sheffield, S1 3JD
[email protected],
[email protected],
[email protected] Abstract. This paper presents the results from collaboration between the National Physical
Laboratory (NPL) and the University of Sheffield on an ongoing research project at NPL. A
50 year old reinforced concrete footbridge has been converted to a full scale structural health
monitoring (SHM) demonstrator. The structure is monitored using a variety of techniques;
however, interrelating results and converting data to knowledge are not possible without a
reliable numerical model. During the first stage of the project, the work concentrated on static
loading and an FE model of the undamaged bridge was created, and updated, under specified
static loading and temperature conditions. This model was found to accurately represent the
response under static loading and it was used to identify locations for sensor installation. The
next stage involves the evaluation of repair/strengthening patches under both static and
dynamic loading. Therefore, before deliberately introducing significant damage, the first set of
dynamic tests was conducted and modal properties were estimated. The measured modal
properties did not match the modal analysis from the statically updated FE model; it was clear
that the existing model required updating. This paper introduces the results of the dynamic
testing and model updating. It is shown that the structure exhibits large non-linear, amplitude
dependant characteristics. This creates a difficult updating process, but we attempt to produce
the best linear representation of the structure. A sensitivity analysis is performed to determine
the most sensitive locations for planned damage/repair scenarios and is used to decide whether
additional sensors will be necessary.
1. Introduction
The footbridge transformed to SHM demonstrator is at the heart of the series of projects run at NPL.
The bridge was in use for over 40 years at NPL. During the demolition of the buildings it connected in
2007, the bridge was moved to its current location suitable for destructive testing. It has become a
truly multidisciplinary collaborative project with more that 40 companies and academic institutions
involved and working together over the last two years. The main aims of the project are to:
• provide test facilities for long-term all year around performance of the sensors covering 18
different technologies employed for continuous monitoring and short-term testing
• demonstrate various advanced real-life monitoring technologies for asset holders and
maintenance companies
• contribute to understanding how to assess the state of the structure and the ways of extending
safe usage of ageing infrastructure.
1
To whom any correspondence should be addressed.
Published under licence by IOP Publishing Ltd 1
9th International Conference on Damage Assessment of Structures (DAMAS 2011) IOP Publishing
Journal of Physics: Conference Series 305 (2011) 012105 doi:10.1088/1742-6596/305/1/012105
More details about the project can be found in [1-2] and summary of technologies shown in figure 1.
Technology Monitoring Measurand contributors
1 Digital Image Survey Full field strain, NPL
Correlation displacement
2 Laser scan Survey Shape John Moore University
3 Optical strain gauge Tests only Displacement Imetrum
technique
4 Optical Fiber Bragg Continuous Strain, RH SmartFibers, City University
Grating
5 Acoustic Emission Continuous Cracking Physical Acoustics, Cardiff
6 Distributed crack sensor Continuous Cracking Strainstall
7 Wireless crack sensor Continuous Crack opening, SPPS
8 Digital leveling Survey Shape SPPS
9 Wireless accelerometer Continuous Tilt SPPS
sensor
10 Resistance strain gauges Continuous Strain Vishay/NPL
11 Vibrating wire sensors Continuous Strain Soil Instruments
12 Electrolevel tilt sensors Continuous Tilt Soil Instruments
13 LVDT Continuous Displacement NPL
14 Corrosion mapping Survey Change in el/magnetic University of Keele, SciSite
field
15 Hydralign Continuous Vertical displacement Pruftechnik
16 Accelerometers Dynamic tests Vibration, mode shape Sheffield University, Full
calculations scale dynamics
17 RFID sensor Inspection Rebar strain Technimeasure
18 Wireless system Continuous New transmission The IMC group
prorochol
Figure 1. A list of technologies used in the bridge project.
This paper presents a case study focusing on updating the FE model for the relatively
undamaged bridge using dynamic vibration tests (modal analysis). A lot of SHM and damage
detection approaches are based on information obtained from dynamic response to identify the
existence, location and magnitude of damage. An accurate and reliable model is important for
assessment and prediction of the performance of the bridge at different stages of damage. Our bridge
has been monitored extensively during the testing program and between the tests. All observed types
of damage, both artificially introduced during tests and naturally occurring due to gradual degradation
of the structure and materials have been recorded. A well-documented history of damage introduced
to the footbridge gives the opportunity to assess the performance of various methods of damage
detection. An experimental program for 2011 will include damage/repair tests with different levels of
damage introduced. The data collected during this period will be used to estimate the sensitivity and
accuracy of various damage detection methods.
2
9th International Conference on Damage Assessment of Structures (DAMAS 2011) IOP Publishing
Journal of Physics: Conference Series 305 (2011) 012105 doi:10.1088/1742-6596/305/1/012105
2. Experiment
2.1 Bridge description
The footbridge-demonstrator, a 15 tonne, 5 metre high, 20 metre long concrete reinforced bridge, is
one of the largest specimens used at NPL. It was built in the mid 1960s using concrete and
reinforcement typical for that time and aged naturally in service for 40 years. Therefore, it has become
a very good example for studying the ageing infrastructure and of the 1950-70s period in particular. It
is a simple construction with a deck supported by two A- frame columns as shown in figure 2.
Figure 2. The bridge in its current location before
installation of any sensors.
2.2 Dynamic vibration tests and modal parameter evaluation
The Vibration Engineering Section (VES), part of the Civil and Structural Engineering Department at
the University of Sheffield, performed an experimental modal analysis (EMA) of the bridge. The
testing consisted of forced vibration testing (FVT) using an electro-dynamic shaker in the vertical,
lateral and longitudinal directions. The test grid used is shown in figure 3. Each point has an
accelerometer measuring either vertically or laterally with the shaker at position 10. A photo of the
setup form the bridge deck is shown in figure 4.
Figure 3. EMA test grid. Figure 4. EMA setup.
3
9th International Conference on Damage Assessment of Structures (DAMAS 2011) IOP Publishing
Journal of Physics: Conference Series 305 (2011) 012105 doi:10.1088/1742-6596/305/1/012105
The FRFs were measured with 50 averages using a 80 Hz sampling rate with a 20 second window. A
sample of the vertical FRFs is shown in figure 5, with a number of clear peaks in the data identifying
well separated modes.
Figure 5. A sample of the vertical measured FRFs.
From curve fitting the FRFs the modal properties were estimated. Figure 6 shows the first 6 vertical
modal parameters ranging from approximately 5 Hz to 32 Hz. The FRFs were very clean and easily
used to estimate the modal properties. The modal mass of mode 5 is much larger than the other modal
masses, indicating that this is a vertical component of a predominantly lateral mode. Although not so
evident in the mode shape plots, but in animated mode shapes, there is clear movement of the
foundations.
Mode Mode Mode
1 2 3
5.1 Hz, 1.89% , 5900kg 8.9 Hz, 1.59%, 3090 kg 12.7 Hz, 2.08% , 1080 kg
Mode Mode Mode
4 5 6
15.5 Hz, 1.85% , 1390 kg 30.4 Hz, 3.2%, 14000kg 31.9 Hz, 5.1% , 2070 kg
Figure 6. First 6 vertical modal properties from the EMA.
During the test some non-linear amplitude dependant behaviour was detected and this was investigated
further. A shaker shutdown test was conducted, which consists of running the shaker at a resonant
frequency and estimating the frequency and damping on a cycle-by-cycle basis of the free decay of the
signal. A number of different frequencies were attempted, but only the excitations at 8.4 Hz and 15
Hz produced a good free decay fit.
Figure 7 shows how frequency and damping varies for each of two frequencies with respect to
the amplitude of acceleration. There is a small change in damping, but the change in frequency is
quite substantial with a large frequency drop with larger amplitudes of vibration.
4
9th International Conference on Damage Assessment of Structures (DAMAS 2011) IOP Publishing
Journal of Physics: Conference Series 305 (2011) 012105 doi:10.1088/1742-6596/305/1/012105
damping vs amplitude; average damping: 1.9192% frequency vs amplitude; average frequency: 8.6072Hz
9.2
2 9
1.5 8.8
f /Hz
ζ /%
1 8.6
0.5 8.4
0 8.2
0 0.5 1 1.5 0 0.5 1 1.5
ch 2 (m/sec2) ch 2 (m/sec2)
damping vs amplitude; average damping: 1.8104% frequency vs amplitude; average frequency: 15.4781Hz
15.8
2
1.5 f /Hz 15.6
ζ /%
1
15.4
0.5
0 15.2
0 0.2 0.4 0.6 0.8 1 0 0.2 0.4 0.6 0.8 1
ch 18 (m/sec2) ch 18 (m/sec2)
d i f l d d i ti
Figure 7 Amplitude dependant behaviour of frequency and damping for 8.4 Hz (top), 15 Hz (bottom)
3. FEA model
An initial three-dimensional linear elastic finite element model of the bridge was constructed where
the geometry of the bridge was created in ABAQUS CAE version 6.9-1. The geometry was meshed
using continuum linear brick elements. As suggested in the ABAQUS manual the C3D8R element
type was chosen for static bending loading (see comparison with Bernoulli-Euler theory prediction in
[3]). The model was validated under static loading and results compared reasonably well with
experiments conducted during static loading tests in 2008-2009. A discussion of the results can be
found in [4].
When vibration analysis was considered, the suitability of the elements was reassessed. Three
types of elements were selected and a comparison of the results using the same material parameters
and boundary conditions showed a significant difference in natural frequencies, see table 1. After an
investigation of different element types the difference was attributed to an effect called sheer locking
or parasitic sheer [5]. The incompatible mode element C3D8I was chosen as the most suitable.for this
work. Elements of the type C3D8I are enhanced by incompatible modes to improve the bending
behaviour to eliminate the parasitic shear stresses and artificial stiffening in bending due to Poisson's
effect [3]. Although a discussion of the full integration, reduced integration, and incompatible
elements is not presented here more details of the effect of different element types and mesh density
on the frequency analysis can be found in [5].
5
9th International Conference on Damage Assessment of Structures (DAMAS 2011) IOP Publishing
Journal of Physics: Conference Series 305 (2011) 012105 doi:10.1088/1742-6596/305/1/012105
Table 1. Summary of natural frequencies obtained from modal analysis for three types of elements
mode Reduced Incompatible Full
integration mode integration
1 5.86 6.13 7.86
2 7.61 7.78 7.96
3 8.63 9.16 10.35
4 9.26 9.63 10.36
5 9.92 10.76 13.10
6 12.30 12.92 14.42
7 15.14 16.02 16.05
8 19.73 20.93 23.99
9 30.27 32.42 36.74
10 34.14 35.68 37.35
The mesh created had 8 elements across the width of the deck, 4 elements through the thickness
and approximately 160 elements along the length. The bases of the foundations of the bridge were
fully constrained. Material properties were applied to the whole bridge to represent concrete plus steel
reinforcement as one single set of properties. The Young’s modulus was 38 kN/mm2, Poisson’s ratio
was 0.2, density was 2.4E-9 tonnes/mm3 (2400 kg/m3) and thermal expansion was 12E-6/°C.
Various loading conditions were analysed. Firstly, a gravity load was applied to the bridge. In
some analyses, the bridge was loaded in the middle of the main span, in other cases cantilever loading
was applied at one end. Secondly, several temperature analyses were performed. In the first scenario,
all nodes in the model were given an initial temperature e.g.0°C. In the first step of the analysis the
temperature of all nodes was increased to a new value e.g. 20°C. The second scenario consisted of
applying a linear temperature gradient along the width of the bridge deck. Finally, to investigate the
modes and natural frequencies of the bridge, a modal analysis perturbation step was added after either
gravity or thermal loading as a pre-stress loading. The lanczos eigensolver was used with the analysis
outputting the first 10 vibration modes of the bridge.
4. Model updating
This case of FE model updating is a first step in our investigation to improve the existing model of the
bridge whilst the bridge remains well within serviceability limits. The next step will be focused on
damage detection after set of damage/ repair experiments are conducted.
The strong amplitude dependency of natural frequencies and damping was observed during the
tests indicate that the frequencies will be not possible to match exactly. A small movement of
foundations were also visible in the EMA mode shapes animation. However, the results obtained from
the FE analysis before updating and EMA agreed reasonably well: the mode shapes looked similar but
there was a small variation in frequencies. Although it is not possible to model amplitude dependant
non-linear variation in modal properties with a linear analysis of a finite element model we have
attempted to use a best linear fit approach discussed below.
4.1 Correlation analysis.
The natural frequencies and modes obtained from the experimental analysis and FE model were
compared. The discussion here will be limited to the 5 vertical modes shown in the tables 2 and 3.
Table 2 shows EMA and FE frequencies, their differences and modal assurance criterion (MAC)
values. Table 3 is a comparison of mode shapes where FE shows all degrees of freedom of the vertical
mode shape, but for EMA only vertical components are presented, thus the columns of the FE model
are clearly moving in the x direction, whereas EMA are not.
6
9th International Conference on Damage Assessment of Structures (DAMAS 2011) IOP Publishing
Journal of Physics: Conference Series 305 (2011) 012105 doi:10.1088/1742-6596/305/1/012105
Table 2. Comparison of natural frequencies of experimental and FE vertical modes
Frequencies Frequencies FE Differences in MAC
EMA frequencies, % values
V1 5.09 4.71 -7.47 0.986
V2 8.92 8.41 -5.72 0.995
V3 12.70 11.97 -5.75 0.991
V4 15.50 16.7 7.74 0.978
V6 31.90 32.68 2.45 0.971
Table 3. Comparison of experimental and FE vertical modes.
EMA experimental FEA model
7
9th International Conference on Damage Assessment of Structures (DAMAS 2011) IOP Publishing
Journal of Physics: Conference Series 305 (2011) 012105 doi:10.1088/1742-6596/305/1/012105
4.2 Sensitivity analysis
Modes obtained from the original FE model (updated by static loading experiments) were close
enough to EMA experimental results to justify a narrow range of input parameters for sensitivity
analysis. Four types of parameters were included in the sensitivity analysis: Young’s modulus,
concrete density, temperature and boundary conditions
As it is assumed that there is no significant localised damage, material properties such as the
Young’s modulus and concrete density can be varied as global parameters in the model and only the
variations in modal frequencies need to be examined.
The effect of temperature was included because it was found that the temperature underwent
substantial daily changes and the variations had a significant influence on the deflection of the bridge
during the static tests (details can be found [4]). Here this effect was studied by comparing the results
of vibration analysis on its own to a model subjected first to temperature changes and then vibration
analysis. Firstly, uniform temperature loading of 20 degrees (equivalent to the daily temperature
variation at the bridge’s location) was applied to the whole structure. Secondly, a linear gradient of 10
degrees, representative of the temperature variation between the open sunny side and shady side of the
bridge near the parkland, was applied across the deck. As the cantilevers of the bridge are not
constrained, the shape of the deck was changed as shown in figure 10. An example of a sensitivity plot
for fixed boundary conditions is shown in figure 11, where D is concrete density, E is global Young’s
modulus and T is temperature. The figure shows that the applied temperature gradient, although
sufficient to change the shape of the deck, has little effect on the frequencies of the vertical modes.
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
V1 D
V2
V3 E
T E D V4
V6 T
Figure 8. Change in shape of the bridge Figure 9. An example of sensitivity plot for fixed
deck due to the temperature gradient of 10 boundary conditions where x-axis is the vertical mode
degrees applied across the deck from front number, y-axis shows model parameter, and z-axis is a
to back. relative sensitivity scale.
A small movement of one of the foundations was observed during the experiment. Boundary
conditions were varied in the model to examine the effect of the constraints of the foundations. It was
found that they are extremely important and vary modes noticeably. These conditions were considered
as fully fixed, free in all directions (x, y and z) with varying stiffness and free in one of the directions
only whilst fixed in other two with varying stiffness. More investigation is needed but some details are
given in section 5.
8
9th International Conference on Damage Assessment of Structures (DAMAS 2011) IOP Publishing
Journal of Physics: Conference Series 305 (2011) 012105 doi:10.1088/1742-6596/305/1/012105
4.3 Model updating
The model was updated using a standard optimisation algorithm to solve a constrained least-squares
problem. The function to be minimised is shown in (1).
5
F = ∑ [ f EMA − f FE ( D, E , k , T )]2 (1)
i =1
where f EMA is a frequency obtained experimentally, D is global concrete density, E is global Young’s
modulus, k stiffness coefficient, T is temperature and f FE is the calculated frequency with parameters
between 2400 kg/m3 < D < 2700 kg/m3, 35 N/mm2 < E < 45 N/mm2 and 0.35 N/m < k < 6*1010 N/m.
5. Results and discussion
5.1 Model updating results
The updated FE frequencies are shown in table 4. The results are also in Figure 5 where FE mode
frequencies for 5 vertical modes plotted against experimental EMA mode frequencies. This is a
straight line if the frequencies match. The set of parameters used are: D=2530 kg/m3, E = 38 N/mm2,
one of the foundations is free in y direction and fixed in x and z ones.
40
FE update frequencies, Hz
Frequencies Frequencies FE Differences in 30
EMA frequencies, %
5.09 4.71 20
V1 -7.47
V2 8.92 8.41 -5.72
10
V3 12.70 11.97 -5.75
V4 15.50 16.7 7.74 0
V6 31.90 32.68 2.45 0 10 20 30 40
EMA frequencies, Hz
Table 4. Comparison of natural frequencies of Figure 5. Frequency correlation between FE
experimental and FE vertical modes. model and experimental modal analysis EMA
after FE model updating.
This work represents a first attempt to use the dynamic testing for the NPL bridge project and a
comparison between experimental and calculated frequencies is very encouraging at this early stage.
It is also important to note that the analysis presented here is semiautomatic. The changes in boundary
conditions (i.e. how free foundations can move) alter the number of modes significantly and non-
linearly and therefore the correlation (section 4.1 above) and sensitivity (section 4.2) analyses must be
repeated for each boundary condition. To overcome this limitation a full three dimensional modal
analysis is required that has sufficient number of modes for optimisation even for the relatively
straightforward geometry of the footbridge.
5.2 Non-linear effects.
Although the structure is simple, the updating task is not trivial, as discussed earlier. Modal analysis is
linear but there are a number of significant non-linearities in the model. The effects of concern are:
• the ground deformation is non-linear, this is currently modelled with a single linear spring – a
more complex method maybe required
• the handrail, although not structural, may add sufficient stiffness to contribute to the overall
stiffness, but due to weak connections, the degree is unknown.
• during the tests we clearly identified a very large amplitude dependant behaviour of the
modes. This is non-linear and hard to model.
To address this situation the dynamic response of the foundations and columns themselves need to be
investigated experimentally in more detail. It is clear from the analysis that the movement of the
9
9th International Conference on Damage Assessment of Structures (DAMAS 2011) IOP Publishing
Journal of Physics: Conference Series 305 (2011) 012105 doi:10.1088/1742-6596/305/1/012105
foundations should be assessed in all three directions. The amount of movement is small and may be
experimentally challenging to monitor accurately. New and more sophisticated optimisation methods
will need to be considered.
5.1 Dynamic testing to improve sensor locations for future damage detection
Determining best sensor locations is an optimisation task on its own and depends on the objectives of
monitoring with different types of constraints, for example access or cost etc. In the current project,
we are mostly interested in getting as much information as possible form the most sensitive locations.
After examining the mode shapes and the positions of nodes it was clear that most of the areas of
interest have already been covered but a new location was identified. Although there are sensors
measuring strain on the foundations on the bottom of the columns, it will be important to use
accelerometers or other sensors to determine the dynamic stiffness of the foundations. Monitoring any
changes during damage accumulation will be important for interpreting the data.
6. Concluding remarks and further work
The results of dynamic tests and modal analysis showed that it is potentially a very powerful tool for
structural assessment, with the modal properties being sensitive to a number of structural parameters.
The FE model updating does not only improve our understanding of the overall behaviour of the
footbridge, but also ranks the sensitivity to different parameters such as material parameters,
temperature and boundary conditions. It will help in preparing for the next stage of the project,
including concrete removal during damage/repair cycles, assessing existing sensor locations and
prediction the variation of modal properties with respect to the induced damage.
Based on this study two new areas will be included in the work plan:
• identify and add more sensors to monitor foundations and columns in more details
• develop more generic and sophisticated optimisation routine to include several substructures
and vary properties of each substructure and boundary conditions.
7. References
[1] Barton E and Zhang B 2010 Proc. Int. Conf Structural Faults & Repair-2010, (Edinburgh) p 20
[2] Barton E and Zhang B 2010 NDE/NDT for Highways and Bridges Structural Materials
Technology (SMT) (New York ASNT)
[3] ABAQUS/CAE User's Manual Version 6.9.
[4] NPL SHM Footbridge Website http://www.npl.co.uk/advanced-materials/materials-
areas/structural-health-monitoring/footbridge-monitoring-project-(shm)
[5] Brown J 1997 MSC Aerospace Users' Conference (MSC Software Corporation)
10
