The aim of this study is the analysis and comparison of several sets of features selected from different domains for the detection of damage induced by scour on historical bridges. A scaled experimental model of a masonry arch bridge was subjected to differential settlements of the pier to simulate the occurrence of scour events. Damage states of increasing extent were introduced to verify the sensitivity of each feature and the accuracy of the damage detection method. The vibration measurements were acquired after each damage step using sensors located in different positions on the structure and different sources of excitation. The Kernel Density Estimation (KDE) was employed to characterise the correlation between the vibration signatures acquired in the time domain. The identified natural frequencies and the sampled range of the transmissibility spectrum were used in the Outlier Analysis (OA) to identify the novelties coming from the damage steps. All the selected features proved to be sensitive to damage, while showing pros and cons depending on the feature domain.
[1]
Charles R. Farrar,et al.
Damage identification and health monitoring of structural and mechanical systems from changes in their vibration characteristics: A literature review
,
1996
.
[2]
Keith Worden,et al.
Non-linearity detection in a masonry arch bridge model subject to artificial settlements
,
2010
.
[3]
Rosario Ceravolo,et al.
Modal Identification of an Experimental Model of Masonry Arch Bridge
,
2009
.
[4]
K. Worden,et al.
The application of machine learning to structural health monitoring
,
2007,
Philosophical Transactions of the Royal Society A: Mathematical, Physical and Engineering Sciences.
[5]
E. Parzen.
On Estimation of a Probability Density Function and Mode
,
1962
.
[6]
Keith Worden,et al.
EXPERIMENTAL VALIDATION OF A STRUCTURAL HEALTH MONITORING METHODOLOGY: PART II. NOVELTY DETECTION ON A GNAT AIRCRAFT
,
2003
.