In this study, a systematic approach was adopted to investigate, numerically
and experimentally, localised defects and/or damage in timber bridges, such as rot,
using modal based damage detection techniques. An existing damage detection method
namely damage index (DI) method that utilises modal strain energy before and after
damaged state was adopted. One contribution of this study was to modify the Dl method
by an additional step of normalising the modal curvature, which would minimise the
dominance of higher modes.
In the numerical models, a comparative study of the effects of numerical
integration techniques used in a damage detection process was carried out. The results
show that when mode shape curvature integrations use the rectangular rule for the
numerical integration, it yields better results than the trapezoidal rule.
In the numerical examples using a finite element model of timber beam, the
modified DI (MDI) methods were found to perform better than its original form for
locating'" single and multiple damage scenarios. For the DI methods, two types of
formulations were adopted and modified, and they are denoted as modified damage
index I (MDI-I) and modified damage index II (MDI-II). Another modal based damage
detection method, namely changes in flexibility (CIF), was adopted for locating
damage. It was found that the ClF method performed reasonably well for single damage
but not multiple damage scenarios.
As part of the study, the modified damage index methods were utilised for
evaluating severity of damage. For the :MDI-I method, the formulation was not derived
to evaluate damage severity directly. Instead, a hybrid of the MDI-I and CIF methods
(HMC), was proposed for evaluating severity of damage in terms of loss of '1' (moment
of inertia). Using three levels of damage, i.e. light (L), medium (M) and severe (S), the
HMC method is able to predict the medium and severe damage quite well, but it is less
efficient for light damage scenarios. For the MDI-II method, further manipulation of the
algorithm can predict the severity of damage in terms of loss of'I'. This method is able
to predict the medium and severe damage quite well but is not as good for the light
damage. Both methods, HMC and MDI-II, for predicting severity of damage, required
some adjustment using a weighting factor in order to obtain reasonable results.
An experimental modal analysis (EMA) test program of timber beams was
undertaken. This was done to verify the robustness of the modified damage index
methods for detecting location and estimating severity of damage. The laboratory
investigation was conducted on the corresponding changes of modal parameters due to
loss of section. The MDI methods were used to detect location of damage and to
evaluate the severity of damage in the test beams. A mode shape reconstruction
technique was utilised to enhance the capability of the damage detection algorithms
with limited number of sensors. The test results and analysis show that location of
damage is quite accurately estimated with the available sensors. The methods
demonstrate that they are less mode dependant and can detect damage with a higher
degree of confidence. The MDI methods also show that they are able to predict the
severe damage well, but it is less accurate for the medium damage and not as good for
The damage index II (DI-II) method extended to plate-like structures (DI-II-P)
was adopted and evaluated for detecting damage. Based on finite element analysis
(FEA) results of a laboratory timber bridge, the DI-II-P method which utilises two dimensional
(2-D) mode shape curvature was employed to detect location of damage.
The results show that the tnethod based on 2-D mode shape curvature is able to locate
damage quite well, numerically. A supplementary work using the DI-II-P method in a
timber plate model was carried out. The results also show that the method was able to
predict the damage location well.
A process of updating a laboratory timber bridge, analytically, is presented. A
finite element model was developed and updated with experimental modal data.
Material properties of timber beam (girders) and plywood (deck) as well as the screw
connection between deck and girder were experimentally investigated. These test results
were then used for the finite element modelling. The model has been developed
sequentially starting with a preliminary model having very simple features. It followed
by the advanced model calibrated with the experimental modal data employing a global
objective function, consisting of errors of natural frequencies and modal assurance
criterion. The calibrated finite element model shows a good correlation to the
experimental model with minor adjustments to the real material properties and boundary
conditions. The calibrated model can reasonably be used to study the damaged
behaviour of the laboratory timber bridge.
The bridge model was then used to verify the numerical results for detecting
damage. The bridge was inflicted with various damage scenarios with loss of section
similar to the timber beam models. The limited number of data was expanded using the
2-D cubic spline. Using the reconstructed data for detecting damage yields better results
than just using 'as is' data. Using the undanlaged and dmnaged modal data, the D I-II -P
method was employed to detect the location of damage. The results of using the first
nine modes showed that generally the severe damage is able to be located by the
method. It performs reasonably well for the medium damage but does not perform as
good in the light damage scenarios. However, in some cases the method can present
some problems in identifying severe damage, which may be due to lack of
normalisation of mode shape curvature. Complementary work was undertaken using the
method 'On a timber plate, experimentally. The results showed that the damage detection
process in the timber plate is less efficient compared to the laboratory timber bridge.
A comprehensive comparative study was carried out based on the results of the
numerical and experimental investigation of damage detection on timber beam,
laboratory timber bridge and timber plate. For the timber beam, both damage detection
methods, MDI-I and MDI-II, were capable of detecting medium and severe damage in
the numerical and experimental studies. However, the light damage was not identified
well using the experimental data in the presence of noise. To estimate damage severity
in the timber beam, the HMC method performed well for the medium and severe
damage. The method did not work well in estimating severity of light damage. Similar
conclusions can be drawn in using the MDI-II method to estimate the damage severity.
The results of applying the DI-II-P method (using 9 modes) to locate damage in the
laboratory timber bridge showed that numerical and experimental data are capable of detecting all severe damage for damage cases with less than three damage locations.
While for light and medium damage, the experimental data did not work well as
compared to the numerical one. For the timber plate (a complementary work), the
numerical and experimental results also showed that they are able to detect the severe
damage well. However, there were serious false positives appearing in the light damage
cases in the experimental results.